Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
authorLinus Torvalds <torvalds@linux-foundation.org>
Thu, 19 Jul 2018 02:32:54 +0000 (19:32 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 19 Jul 2018 02:32:54 +0000 (19:32 -0700)
Pull networking fixes from David Miller:
 "Lots of fixes, here goes:

   1) NULL deref in qtnfmac, from Gustavo A. R. Silva.

   2) Kernel oops when fw download fails in rtlwifi, from Ping-Ke Shih.

   3) Lost completion messages in AF_XDP, from Magnus Karlsson.

   4) Correct bogus self-assignment in rhashtable, from Rishabh
      Bhatnagar.

   5) Fix regression in ipv6 route append handling, from David Ahern.

   6) Fix masking in __set_phy_supported(), from Heiner Kallweit.

   7) Missing module owner set in x_tables icmp, from Florian Westphal.

   8) liquidio's timeouts are HZ dependent, fix from Nicholas Mc Guire.

   9) Link setting fixes for sh_eth and ravb, from Vladimir Zapolskiy.

  10) Fix NULL deref when using chains in act_csum, from Davide Caratti.

  11) XDP_REDIRECT needs to check if the interface is up and whether the
      MTU is sufficient. From Toshiaki Makita.

  12) Net diag can do a double free when killing TCP_NEW_SYN_RECV
      connections, from Lorenzo Colitti.

  13) nf_defrag in ipv6 can unnecessarily hold onto dst entries for a
      full minute, delaying device unregister. From Eric Dumazet.

  14) Update MAC entries in the correct order in ixgbe, from Alexander
      Duyck.

  15) Don't leave partial mangles bpf program in jit_subprogs, from
      Daniel Borkmann.

  16) Fix pfmemalloc SKB state propagation, from Stefano Brivio.

  17) Fix ACK handling in DCTCP congestion control, from Yuchung Cheng.

  18) Use after free in tun XDP_TX, from Toshiaki Makita.

  19) Stale ipv6 header pointer in ipv6 gre code, from Prashant Bhole.

  20) Don't reuse remainder of RX page when XDP is set in mlx4, from
      Saeed Mahameed.

  21) Fix window probe handling of TCP rapair sockets, from Stefan
      Baranoff.

  22) Missing socket locking in smc_ioctl(), from Ursula Braun.

  23) IPV6_ILA needs DST_CACHE, from Arnd Bergmann.

  24) Spectre v1 fix in cxgb3, from Gustavo A. R. Silva.

  25) Two spots in ipv6 do a rol32() on a hash value but ignore the
      result. Fixes from Colin Ian King"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (176 commits)
  tcp: identify cryptic messages as TCP seq # bugs
  ptp: fix missing break in switch
  hv_netvsc: Fix napi reschedule while receive completion is busy
  MAINTAINERS: Drop inactive Vitaly Bordug's email
  net: cavium: Add fine-granular dependencies on PCI
  net: qca_spi: Fix log level if probe fails
  net: qca_spi: Make sure the QCA7000 reset is triggered
  net: qca_spi: Avoid packet drop during initial sync
  ipv6: fix useless rol32 call on hash
  ipv6: sr: fix useless rol32 call on hash
  net: sched: Using NULL instead of plain integer
  net: usb: asix: replace mii_nway_restart in resume path
  net: cxgb3_main: fix potential Spectre v1
  lib/rhashtable: consider param->min_size when setting initial table size
  net/smc: reset recv timeout after clc handshake
  net/smc: add error handling for get_user()
  net/smc: optimize consumer cursor updates
  net/nfc: Avoid stalls when nfc_alloc_send_skb() returned NULL.
  ipv6: ila: select CONFIG_DST_CACHE
  net: usb: rtl8150: demote allmulti message to dev_dbg()
  ...

171 files changed:
Documentation/networking/bonding.txt
Documentation/networking/e100.rst
Documentation/networking/e1000.rst
MAINTAINERS
drivers/net/ethernet/aquantia/atlantic/aq_cfg.h
drivers/net/ethernet/aquantia/atlantic/aq_hw.h
drivers/net/ethernet/aquantia/atlantic/aq_main.c
drivers/net/ethernet/aquantia/atlantic/aq_nic.c
drivers/net/ethernet/aquantia/atlantic/aq_nic.h
drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c
drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c
drivers/net/ethernet/broadcom/bcmsysport.c
drivers/net/ethernet/broadcom/bcmsysport.h
drivers/net/ethernet/broadcom/bnxt/bnxt.c
drivers/net/ethernet/broadcom/bnxt/bnxt.h
drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c
drivers/net/ethernet/broadcom/bnxt/bnxt_ulp.c
drivers/net/ethernet/broadcom/tg3.c
drivers/net/ethernet/broadcom/tg3.h
drivers/net/ethernet/cadence/macb.h
drivers/net/ethernet/cadence/macb_main.c
drivers/net/ethernet/cavium/Kconfig
drivers/net/ethernet/cavium/liquidio/lio_main.c
drivers/net/ethernet/cavium/octeon/octeon_mgmt.c
drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c
drivers/net/ethernet/chelsio/cxgb4/t4_hw.c
drivers/net/ethernet/ibm/ibmvnic.c
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
drivers/net/ethernet/mellanox/mlx4/en_rx.c
drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c
drivers/net/ethernet/qlogic/qed/qed.h
drivers/net/ethernet/qlogic/qed/qed_debug.c
drivers/net/ethernet/qlogic/qed/qed_main.c
drivers/net/ethernet/qlogic/qed/qed_mcp.c
drivers/net/ethernet/qlogic/qlcnic/qlcnic_sysfs.c
drivers/net/ethernet/qualcomm/qca_spi.c
drivers/net/ethernet/realtek/r8169.c
drivers/net/ethernet/renesas/ravb_main.c
drivers/net/ethernet/renesas/sh_eth.c
drivers/net/ethernet/sfc/ef10.c
drivers/net/ethernet/sfc/efx.c
drivers/net/ethernet/stmicro/stmmac/dwmac-sun8i.c
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c
drivers/net/hyperv/netvsc.c
drivers/net/hyperv/rndis_filter.c
drivers/net/ieee802154/adf7242.c
drivers/net/ieee802154/at86rf230.c
drivers/net/ieee802154/fakelb.c
drivers/net/ieee802154/mcr20a.c
drivers/net/phy/marvell.c
drivers/net/phy/phy_device.c
drivers/net/phy/sfp-bus.c
drivers/net/tun.c
drivers/net/usb/asix_devices.c
drivers/net/usb/lan78xx.c
drivers/net/usb/qmi_wwan.c
drivers/net/usb/rtl8150.c
drivers/net/usb/smsc75xx.c
drivers/net/wireless/ath/ath10k/mac.c
drivers/net/wireless/ath/ath10k/wmi.h
drivers/net/wireless/ath/wcn36xx/testmode.c
drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c
drivers/net/wireless/marvell/mwifiex/usb.c
drivers/net/wireless/mediatek/mt7601u/phy.c
drivers/net/wireless/quantenna/qtnfmac/cfg80211.c
drivers/net/wireless/realtek/rtlwifi/base.c
drivers/net/wireless/realtek/rtlwifi/base.h
drivers/net/wireless/realtek/rtlwifi/core.c
drivers/net/wireless/realtek/rtlwifi/pci.c
drivers/net/wireless/realtek/rtlwifi/ps.c
drivers/net/wireless/realtek/rtlwifi/usb.c
drivers/ptp/ptp_chardev.c
include/linux/bpf-cgroup.h
include/linux/filter.h
include/linux/fsl/guts.h
include/linux/if_bridge.h
include/linux/igmp.h
include/linux/marvell_phy.h
include/linux/skbuff.h
include/net/ip6_route.h
include/net/ipv6.h
include/net/netfilter/nf_tables_core.h
include/net/netfilter/nf_tproxy.h
include/net/tc_act/tc_csum.h
include/net/tc_act/tc_tunnel_key.h
include/net/tcp.h
include/net/xdp_sock.h
include/uapi/linux/ethtool.h
include/uapi/linux/tcp.h
kernel/bpf/btf.c
kernel/bpf/devmap.c
kernel/bpf/hashtab.c
kernel/bpf/sockmap.c
kernel/bpf/syscall.c
kernel/bpf/verifier.c
lib/rhashtable.c
net/batman-adv/bat_iv_ogm.c
net/batman-adv/bat_v.c
net/batman-adv/debugfs.c
net/batman-adv/debugfs.h
net/batman-adv/hard-interface.c
net/batman-adv/translation-table.c
net/bpf/test_run.c
net/core/filter.c
net/core/gen_stats.c
net/core/skbuff.c
net/dns_resolver/dns_key.c
net/ieee802154/6lowpan/core.c
net/ipv4/fib_frontend.c
net/ipv4/igmp.c
net/ipv4/inet_fragment.c
net/ipv4/ip_sockglue.c
net/ipv4/netfilter/ip_tables.c
net/ipv4/netfilter/nf_tproxy_ipv4.c
net/ipv4/sysctl_net_ipv4.c
net/ipv4/tcp.c
net/ipv4/tcp_dctcp.c
net/ipv4/tcp_ipv4.c
net/ipv4/tcp_output.c
net/ipv6/Kconfig
net/ipv6/calipso.c
net/ipv6/exthdrs.c
net/ipv6/ip6_fib.c
net/ipv6/ip6_gre.c
net/ipv6/ipv6_sockglue.c
net/ipv6/mcast.c
net/ipv6/ndisc.c
net/ipv6/netfilter/ip6_tables.c
net/ipv6/netfilter/nf_conntrack_reasm.c
net/ipv6/netfilter/nf_tproxy_ipv6.c
net/ipv6/route.c
net/ipv6/seg6_iptunnel.c
net/netfilter/Kconfig
net/netfilter/Makefile
net/netfilter/nf_conntrack_core.c
net/netfilter/nf_tables_set_core.c [new file with mode: 0644]
net/netfilter/nft_compat.c
net/netfilter/nft_set_bitmap.c
net/netfilter/nft_set_hash.c
net/netfilter/nft_set_rbtree.c
net/netfilter/xt_TPROXY.c
net/nfc/llcp_commands.c
net/nsh/nsh.c
net/packet/af_packet.c
net/qrtr/qrtr.c
net/sched/act_csum.c
net/sched/act_tunnel_key.c
net/sched/cls_api.c
net/sched/sch_fq_codel.c
net/sctp/transport.c
net/smc/af_smc.c
net/smc/smc_clc.c
net/smc/smc_close.c
net/smc/smc_tx.c
net/tipc/discover.c
net/tipc/net.c
net/tipc/node.c
net/tls/tls_sw.c
net/xdp/xsk.c
net/xdp/xsk_queue.h
samples/bpf/.gitignore [new file with mode: 0644]
samples/bpf/parse_varlen.c
samples/bpf/test_overhead_user.c
samples/bpf/trace_event_user.c
samples/bpf/xdp2skb_meta.sh
samples/bpf/xdpsock_user.c
scripts/tags.sh
tools/testing/selftests/bpf/test_verifier.c
tools/testing/selftests/net/fib_tests.sh
tools/testing/selftests/net/udpgso_bench.sh

index c13214d073a4866f49025033a86fed03275bca5f..d3e5dd26db12d75bc09d25cacbf0f775003cd527 100644 (file)
@@ -1490,7 +1490,7 @@ To remove an ARP target:
 
 To configure the interval between learning packet transmits:
 # echo 12 > /sys/class/net/bond0/bonding/lp_interval
-       NOTE: the lp_inteval is the number of seconds between instances where
+       NOTE: the lp_interval is the number of seconds between instances where
 the bonding driver sends learning packets to each slaves peer switch.  The
 default interval is 1 second.
 
index 9708f5fa76de2deae7ddf562e757c3d9c4e90079..f81111eba9c5dd157aecdd1f1370031d3d1b1f0f 100644 (file)
@@ -47,41 +47,45 @@ Driver Configuration Parameters
 The default value for each parameter is generally the recommended setting,
 unless otherwise noted.
 
-Rx Descriptors: Number of receive descriptors. A receive descriptor is a data
+Rx Descriptors:
+   Number of receive descriptors. A receive descriptor is a data
    structure that describes a receive buffer and its attributes to the network
    controller. The data in the descriptor is used by the controller to write
    data from the controller to host memory. In the 3.x.x driver the valid range
    for this parameter is 64-256. The default value is 256. This parameter can be
    changed using the command::
 
-   ethtool -G eth? rx n
+     ethtool -G eth? rx n
 
    Where n is the number of desired Rx descriptors.
 
-Tx Descriptors: Number of transmit descriptors. A transmit descriptor is a data
+Tx Descriptors:
+   Number of transmit descriptors. A transmit descriptor is a data
    structure that describes a transmit buffer and its attributes to the network
    controller. The data in the descriptor is used by the controller to read
    data from the host memory to the controller. In the 3.x.x driver the valid
    range for this parameter is 64-256. The default value is 128. This parameter
    can be changed using the command::
 
-   ethtool -G eth? tx n
+     ethtool -G eth? tx n
 
    Where n is the number of desired Tx descriptors.
 
-Speed/Duplex: The driver auto-negotiates the link speed and duplex settings by
+Speed/Duplex:
+   The driver auto-negotiates the link speed and duplex settings by
    default. The ethtool utility can be used as follows to force speed/duplex.::
 
-   ethtool -s eth?  autoneg off speed {10|100} duplex {full|half}
+     ethtool -s eth?  autoneg off speed {10|100} duplex {full|half}
 
    NOTE: setting the speed/duplex to incorrect values will cause the link to
    fail.
 
-Event Log Message Level:  The driver uses the message level flag to log events
+Event Log Message Level:
+   The driver uses the message level flag to log events
    to syslog. The message level can be set at driver load time. It can also be
    set using the command::
 
-   ethtool -s eth? msglvl n
+     ethtool -s eth? msglvl n
 
 
 Additional Configurations
@@ -92,7 +96,7 @@ Configuring the Driver on Different Distributions
 
 Configuring a network driver to load properly when the system is started
 is distribution dependent.  Typically, the configuration process involves
-adding an alias line to /etc/modprobe.d/*.conf as well as editing other
+adding an alias line to `/etc/modprobe.d/*.conf` as well as editing other
 system startup scripts and/or configuration files.  Many popular Linux
 distributions ship with tools to make these changes for you.  To learn
 the proper way to configure a network device for your system, refer to
@@ -160,7 +164,10 @@ This results in unbalanced receive traffic.
 If you have multiple interfaces in a server, either turn on ARP
 filtering by
 
-(1) entering:: echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter
+(1) entering::
+
+       echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter
+
     (this only works if your kernel's version is higher than 2.4.5), or
 
 (2) installing the interfaces in separate broadcast domains (either
index 144b87eef15341059b06655dcc1b0b65c5822b2a..f10dd40869218cb11e1d29bc5e4b6431d30af946 100644 (file)
@@ -34,7 +34,8 @@ Command Line Parameters
 The default value for each parameter is generally the recommended setting,
 unless otherwise noted.
 
-NOTES:  For more information about the AutoNeg, Duplex, and Speed
+NOTES:
+       For more information about the AutoNeg, Duplex, and Speed
         parameters, see the "Speed and Duplex Configuration" section in
         this document.
 
@@ -45,22 +46,27 @@ NOTES:  For more information about the AutoNeg, Duplex, and Speed
 
 AutoNeg
 -------
+
 (Supported only on adapters with copper connections)
-Valid Range:   0x01-0x0F, 0x20-0x2F
-Default Value: 0x2F
+
+:Valid Range:   0x01-0x0F, 0x20-0x2F
+:Default Value: 0x2F
 
 This parameter is a bit-mask that specifies the speed and duplex settings
 advertised by the adapter.  When this parameter is used, the Speed and
 Duplex parameters must not be specified.
 
-NOTE:  Refer to the Speed and Duplex section of this readme for more
+NOTE:
+       Refer to the Speed and Duplex section of this readme for more
        information on the AutoNeg parameter.
 
 Duplex
 ------
+
 (Supported only on adapters with copper connections)
-Valid Range:   0-2 (0=auto-negotiate, 1=half, 2=full)
-Default Value: 0
+
+:Valid Range:   0-2 (0=auto-negotiate, 1=half, 2=full)
+:Default Value: 0
 
 This defines the direction in which data is allowed to flow.  Can be
 either one or two-directional.  If both Duplex and the link partner are
@@ -70,18 +76,22 @@ duplex.
 
 FlowControl
 -----------
-Valid Range:   0-3 (0=none, 1=Rx only, 2=Tx only, 3=Rx&Tx)
-Default Value: Reads flow control settings from the EEPROM
+
+:Valid Range:   0-3 (0=none, 1=Rx only, 2=Tx only, 3=Rx&Tx)
+:Default Value: Reads flow control settings from the EEPROM
 
 This parameter controls the automatic generation(Tx) and response(Rx)
 to Ethernet PAUSE frames.
 
 InterruptThrottleRate
 ---------------------
+
 (not supported on Intel(R) 82542, 82543 or 82544-based adapters)
-Valid Range:   0,1,3,4,100-100000 (0=off, 1=dynamic, 3=dynamic conservative,
-                                 4=simplified balancing)
-Default Value: 3
+
+:Valid Range:
+   0,1,3,4,100-100000 (0=off, 1=dynamic, 3=dynamic conservative,
+   4=simplified balancing)
+:Default Value: 3
 
 The driver can limit the amount of interrupts per second that the adapter
 will generate for incoming packets. It does this by writing a value to the
@@ -135,13 +145,15 @@ Setting InterruptThrottleRate to 0 turns off any interrupt moderation
 and may improve small packet latency, but is generally not suitable
 for bulk throughput traffic.
 
-NOTE:  InterruptThrottleRate takes precedence over the TxAbsIntDelay and
+NOTE:
+       InterruptThrottleRate takes precedence over the TxAbsIntDelay and
        RxAbsIntDelay parameters.  In other words, minimizing the receive
        and/or transmit absolute delays does not force the controller to
        generate more interrupts than what the Interrupt Throttle Rate
        allows.
 
-CAUTION:  If you are using the Intel(R) PRO/1000 CT Network Connection
+CAUTION:
+          If you are using the Intel(R) PRO/1000 CT Network Connection
           (controller 82547), setting InterruptThrottleRate to a value
           greater than 75,000, may hang (stop transmitting) adapters
           under certain network conditions.  If this occurs a NETDEV
@@ -151,7 +163,8 @@ CAUTION:  If you are using the Intel(R) PRO/1000 CT Network Connection
           hang, ensure that InterruptThrottleRate is set no greater
           than 75,000 and is not set to 0.
 
-NOTE:  When e1000 is loaded with default settings and multiple adapters
+NOTE:
+       When e1000 is loaded with default settings and multiple adapters
        are in use simultaneously, the CPU utilization may increase non-
        linearly.  In order to limit the CPU utilization without impacting
        the overall throughput, we recommend that you load the driver as
@@ -168,9 +181,11 @@ NOTE:  When e1000 is loaded with default settings and multiple adapters
 
 RxDescriptors
 -------------
-Valid Range:   48-256 for 82542 and 82543-based adapters
-               48-4096 for all other supported adapters
-Default Value: 256
+
+:Valid Range:
+ - 48-256 for 82542 and 82543-based adapters
+ - 48-4096 for all other supported adapters
+:Default Value: 256
 
 This value specifies the number of receive buffer descriptors allocated
 by the driver.  Increasing this value allows the driver to buffer more
@@ -180,15 +195,17 @@ Each descriptor is 16 bytes.  A receive buffer is also allocated for each
 descriptor and can be either 2048, 4096, 8192, or 16384 bytes, depending
 on the MTU setting. The maximum MTU size is 16110.
 
-NOTE:  MTU designates the frame size.  It only needs to be set for Jumbo
+NOTE:
+       MTU designates the frame size.  It only needs to be set for Jumbo
        Frames.  Depending on the available system resources, the request
        for a higher number of receive descriptors may be denied.  In this
        case, use a lower number.
 
 RxIntDelay
 ----------
-Valid Range:   0-65535 (0=off)
-Default Value: 0
+
+:Valid Range:   0-65535 (0=off)
+:Default Value: 0
 
 This value delays the generation of receive interrupts in units of 1.024
 microseconds.  Receive interrupt reduction can improve CPU efficiency if
@@ -198,7 +215,8 @@ of TCP traffic.  If the system is reporting dropped receives, this value
 may be set too high, causing the driver to run out of available receive
 descriptors.
 
-CAUTION:  When setting RxIntDelay to a value other than 0, adapters may
+CAUTION:
+          When setting RxIntDelay to a value other than 0, adapters may
           hang (stop transmitting) under certain network conditions.  If
           this occurs a NETDEV WATCHDOG message is logged in the system
           event log.  In addition, the controller is automatically reset,
@@ -207,9 +225,11 @@ CAUTION:  When setting RxIntDelay to a value other than 0, adapters may
 
 RxAbsIntDelay
 -------------
+
 (This parameter is supported only on 82540, 82545 and later adapters.)
-Valid Range:   0-65535 (0=off)
-Default Value: 128
+
+:Valid Range:   0-65535 (0=off)
+:Default Value: 128
 
 This value, in units of 1.024 microseconds, limits the delay in which a
 receive interrupt is generated.  Useful only if RxIntDelay is non-zero,
@@ -220,9 +240,11 @@ conditions.
 
 Speed
 -----
+
 (This parameter is supported only on adapters with copper connections.)
-Valid Settings: 0, 10, 100, 1000
-Default Value:  0 (auto-negotiate at all supported speeds)
+
+:Valid Settings: 0, 10, 100, 1000
+:Default Value:  0 (auto-negotiate at all supported speeds)
 
 Speed forces the line speed to the specified value in megabits per second
 (Mbps).  If this parameter is not specified or is set to 0 and the link
@@ -231,22 +253,26 @@ speed.  Duplex should also be set when Speed is set to either 10 or 100.
 
 TxDescriptors
 -------------
-Valid Range:   48-256 for 82542 and 82543-based adapters
-               48-4096 for all other supported adapters
-Default Value: 256
+
+:Valid Range:
+  - 48-256 for 82542 and 82543-based adapters
+  - 48-4096 for all other supported adapters
+:Default Value: 256
 
 This value is the number of transmit descriptors allocated by the driver.
 Increasing this value allows the driver to queue more transmits.  Each
 descriptor is 16 bytes.
 
-NOTE:  Depending on the available system resources, the request for a
+NOTE:
+       Depending on the available system resources, the request for a
        higher number of transmit descriptors may be denied.  In this case,
        use a lower number.
 
 TxIntDelay
 ----------
-Valid Range:   0-65535 (0=off)
-Default Value: 8
+
+:Valid Range:   0-65535 (0=off)
+:Default Value: 8
 
 This value delays the generation of transmit interrupts in units of
 1.024 microseconds.  Transmit interrupt reduction can improve CPU
@@ -256,9 +282,11 @@ causing the driver to run out of available transmit descriptors.
 
 TxAbsIntDelay
 -------------
+
 (This parameter is supported only on 82540, 82545 and later adapters.)
-Valid Range:   0-65535 (0=off)
-Default Value: 32
+
+:Valid Range:   0-65535 (0=off)
+:Default Value: 32
 
 This value, in units of 1.024 microseconds, limits the delay in which a
 transmit interrupt is generated.  Useful only if TxIntDelay is non-zero,
@@ -269,18 +297,21 @@ network conditions.
 
 XsumRX
 ------
+
 (This parameter is NOT supported on the 82542-based adapter.)
-Valid Range:   0-1
-Default Value: 1
+
+:Valid Range:   0-1
+:Default Value: 1
 
 A value of '1' indicates that the driver should enable IP checksum
 offload for received packets (both UDP and TCP) to the adapter hardware.
 
 Copybreak
 ---------
-Valid Range:   0-xxxxxxx (0=off)
-Default Value: 256
-Usage: modprobe e1000.ko copybreak=128
+
+:Valid Range:   0-xxxxxxx (0=off)
+:Default Value: 256
+:Usage: modprobe e1000.ko copybreak=128
 
 Driver copies all packets below or equaling this size to a fresh RX
 buffer before handing it up the stack.
@@ -292,8 +323,9 @@ it is also available during runtime at
 
 SmartPowerDownEnable
 --------------------
-Valid Range: 0-1
-Default Value:  0 (disabled)
+
+:Valid Range: 0-1
+:Default Value:  0 (disabled)
 
 Allows PHY to turn off in lower power states. The user can turn off
 this parameter in supported chipsets.
@@ -309,14 +341,14 @@ fiber interface board only links at 1000 Mbps full-duplex.
 
 For copper-based boards, the keywords interact as follows:
 
-  The default operation is auto-negotiate.  The board advertises all
+- The default operation is auto-negotiate.  The board advertises all
   supported speed and duplex combinations, and it links at the highest
   common speed and duplex mode IF the link partner is set to auto-negotiate.
 
-  If Speed = 1000, limited auto-negotiation is enabled and only 1000 Mbps
+- If Speed = 1000, limited auto-negotiation is enabled and only 1000 Mbps
   is advertised (The 1000BaseT spec requires auto-negotiation.)
 
-  If Speed = 10 or 100, then both Speed and Duplex should be set.  Auto-
+- If Speed = 10 or 100, then both Speed and Duplex should be set.  Auto-
   negotiation is disabled, and the AutoNeg parameter is ignored.  Partner
   SHOULD also be forced.
 
@@ -328,13 +360,15 @@ process.
 The parameter may be specified as either a decimal or hexadecimal value as
 determined by the bitmap below.
 
+============== ====== ====== ======= ======= ====== ====== ======= ======
 Bit position   7      6      5       4       3      2      1       0
 Decimal Value  128    64     32      16      8      4      2       1
 Hex value      80     40     20      10      8      4      2       1
 Speed (Mbps)   N/A    N/A    1000    N/A     100    100    10      10
 Duplex                       Full            Full   Half   Full    Half
+============== ====== ====== ======= ======= ====== ====== ======= ======
 
-Some examples of using AutoNeg:
+Some examples of using AutoNeg::
 
   modprobe e1000 AutoNeg=0x01 (Restricts autonegotiation to 10 Half)
   modprobe e1000 AutoNeg=1 (Same as above)
@@ -357,56 +391,59 @@ Additional Configurations
 
 Jumbo Frames
 ------------
-Jumbo Frames support is enabled by changing the MTU to a value larger
-than the default of 1500.  Use the ifconfig command to increase the MTU
-size.  For example::
+
+  Jumbo Frames support is enabled by changing the MTU to a value larger than
+  the default of 1500.  Use the ifconfig command to increase the MTU size.
+  For example::
 
        ifconfig eth<x> mtu 9000 up
 
-This setting is not saved across reboots.  It can be made permanent if
-you add::
+  This setting is not saved across reboots.  It can be made permanent if
+  you add::
 
        MTU=9000
 
-to the file /etc/sysconfig/network-scripts/ifcfg-eth<x>.  This example
-applies to the Red Hat distributions; other distributions may store this
-setting in a different location.
+  to the file /etc/sysconfig/network-scripts/ifcfg-eth<x>.  This example
+  applies to the Red Hat distributions; other distributions may store this
+  setting in a different location.
+
+Notes:
+  Degradation in throughput performance may be observed in some Jumbo frames
+  environments. If this is observed, increasing the application's socket buffer
+  size and/or increasing the /proc/sys/net/ipv4/tcp_*mem entry values may help.
+  See the specific application manual and /usr/src/linux*/Documentation/
+  networking/ip-sysctl.txt for more details.
 
-Notes: Degradation in throughput performance may be observed in some
-Jumbo frames environments.  If this is observed, increasing the
-application's socket buffer size and/or increasing the
-/proc/sys/net/ipv4/tcp_*mem entry values may help.  See the specific
-application manual and /usr/src/linux*/Documentation/
-networking/ip-sysctl.txt for more details.
+  - The maximum MTU setting for Jumbo Frames is 16110.  This value coincides
+    with the maximum Jumbo Frames size of 16128.
 
-- The maximum MTU setting for Jumbo Frames is 16110.  This value
-  coincides with the maximum Jumbo Frames size of 16128.
+  - Using Jumbo frames at 10 or 100 Mbps is not supported and may result in
+    poor performance or loss of link.
 
-- Using Jumbo frames at 10 or 100 Mbps is not supported and may result
-  in poor performance or loss of link.
+  - Adapters based on the Intel(R) 82542 and 82573V/E controller do not
+    support Jumbo Frames. These correspond to the following product names::
 
-- Adapters based on the Intel(R) 82542 and 82573V/E controller do not
-  support Jumbo Frames.  These correspond to the following product names:
-  Intel(R) PRO/1000 Gigabit Server Adapter Intel(R) PRO/1000 PM Network
-  Connection
+     Intel(R) PRO/1000 Gigabit Server Adapter
+     Intel(R) PRO/1000 PM Network Connection
 
 ethtool
 -------
-The driver utilizes the ethtool interface for driver configuration and
-diagnostics, as well as displaying statistical information.  The ethtool
-version 1.6 or later is required for this functionality.
 
-The latest release of ethtool can be found from
-https://www.kernel.org/pub/software/network/ethtool/
+  The driver utilizes the ethtool interface for driver configuration and
+  diagnostics, as well as displaying statistical information.  The ethtool
+  version 1.6 or later is required for this functionality.
+
+  The latest release of ethtool can be found from
+  https://www.kernel.org/pub/software/network/ethtool/
 
 Enabling Wake on LAN* (WoL)
 ---------------------------
-WoL is configured through the ethtool* utility.
 
-WoL will be enabled on the system during the next shut down or reboot.
-For this driver version, in order to enable WoL, the e1000 driver must be
-loaded when shutting down or rebooting the system.
+  WoL is configured through the ethtool* utility.
 
+  WoL will be enabled on the system during the next shut down or reboot.
+  For this driver version, in order to enable WoL, the e1000 driver must be
+  loaded when shutting down or rebooting the system.
 
 Support
 =======
index 192d7f73fd017255d0697bad499244655463b366..1505c8ea8e7b2e44126bd92f459c2f26fb4dd7c8 100644 (file)
@@ -2523,7 +2523,7 @@ S:        Supported
 F:     drivers/scsi/esas2r
 
 ATUSB IEEE 802.15.4 RADIO DRIVER
-M:     Stefan Schmidt <stefan@osg.samsung.com>
+M:     Stefan Schmidt <stefan@datenfreihafen.org>
 L:     linux-wpan@vger.kernel.org
 S:     Maintained
 F:     drivers/net/ieee802154/atusb.c
@@ -5790,7 +5790,6 @@ F:        include/linux/fsl/
 
 FREESCALE SOC FS_ENET DRIVER
 M:     Pantelis Antoniou <pantelis.antoniou@gmail.com>
-M:     Vitaly Bordug <vbordug@ru.mvista.com>
 L:     linuxppc-dev@lists.ozlabs.org
 L:     netdev@vger.kernel.org
 S:     Maintained
@@ -6909,7 +6908,7 @@ F:        drivers/clk/clk-versaclock5.c
 
 IEEE 802.15.4 SUBSYSTEM
 M:     Alexander Aring <alex.aring@gmail.com>
-M:     Stefan Schmidt <stefan@osg.samsung.com>
+M:     Stefan Schmidt <stefan@datenfreihafen.org>
 L:     linux-wpan@vger.kernel.org
 W:     http://wpan.cakelab.org/
 T:     git git://git.kernel.org/pub/scm/linux/kernel/git/sschmidt/wpan.git
@@ -8629,7 +8628,7 @@ MARVELL MWIFIEX WIRELESS DRIVER
 M:     Amitkumar Karwar <amitkarwar@gmail.com>
 M:     Nishant Sarmukadam <nishants@marvell.com>
 M:     Ganapathi Bhat <gbhat@marvell.com>
-M:     Xinming Hu <huxm@marvell.com>
+M:     Xinming Hu <huxinming820@gmail.com>
 L:     linux-wireless@vger.kernel.org
 S:     Maintained
 F:     drivers/net/wireless/marvell/mwifiex/
index fc7383106946ca6461f62ea305be0f03bb59c227..91eb8910b1c992b1b7876f05a26753a5cf79c100 100644 (file)
@@ -63,8 +63,6 @@
 
 #define AQ_CFG_NAPI_WEIGHT     64U
 
-#define AQ_CFG_MULTICAST_ADDRESS_MAX     32U
-
 /*#define AQ_CFG_MAC_ADDR_PERMANENT {0x30, 0x0E, 0xE3, 0x12, 0x34, 0x56}*/
 
 #define AQ_NIC_FC_OFF    0U
index a2d416b24ffc251c71d002a9befe825d5c585fbc..2c6ebd91a9f2782e87472e497447b60974a7a571 100644 (file)
@@ -98,6 +98,8 @@ struct aq_stats_s {
 #define AQ_HW_MEDIA_TYPE_TP    1U
 #define AQ_HW_MEDIA_TYPE_FIBRE 2U
 
+#define AQ_HW_MULTICAST_ADDRESS_MAX     32U
+
 struct aq_hw_s {
        atomic_t flags;
        u8 rbl_enabled:1;
@@ -177,7 +179,7 @@ struct aq_hw_ops {
                                    unsigned int packet_filter);
 
        int (*hw_multicast_list_set)(struct aq_hw_s *self,
-                                    u8 ar_mac[AQ_CFG_MULTICAST_ADDRESS_MAX]
+                                    u8 ar_mac[AQ_HW_MULTICAST_ADDRESS_MAX]
                                     [ETH_ALEN],
                                     u32 count);
 
index ba5fe8c4125d85d0050c2c5b269cba2c320d3127..e3ae29e523f0e26738b0ab80a2f3ac431083d13b 100644 (file)
@@ -135,17 +135,10 @@ err_exit:
 static void aq_ndev_set_multicast_settings(struct net_device *ndev)
 {
        struct aq_nic_s *aq_nic = netdev_priv(ndev);
-       int err = 0;
 
-       err = aq_nic_set_packet_filter(aq_nic, ndev->flags);
-       if (err < 0)
-               return;
+       aq_nic_set_packet_filter(aq_nic, ndev->flags);
 
-       if (netdev_mc_count(ndev)) {
-               err = aq_nic_set_multicast_list(aq_nic, ndev);
-               if (err < 0)
-                       return;
-       }
+       aq_nic_set_multicast_list(aq_nic, ndev);
 }
 
 static const struct net_device_ops aq_ndev_ops = {
index 1a1a6380c128c4522b330907cc16258f0e012189..7a22d0257e04ccf07ef87cae18d5d4f87630660a 100644 (file)
@@ -563,34 +563,41 @@ err_exit:
 
 int aq_nic_set_multicast_list(struct aq_nic_s *self, struct net_device *ndev)
 {
+       unsigned int packet_filter = self->packet_filter;
        struct netdev_hw_addr *ha = NULL;
        unsigned int i = 0U;
 
-       self->mc_list.count = 0U;
-
-       netdev_for_each_mc_addr(ha, ndev) {
-               ether_addr_copy(self->mc_list.ar[i++], ha->addr);
-               ++self->mc_list.count;
+       self->mc_list.count = 0;
+       if (netdev_uc_count(ndev) > AQ_HW_MULTICAST_ADDRESS_MAX) {
+               packet_filter |= IFF_PROMISC;
+       } else {
+               netdev_for_each_uc_addr(ha, ndev) {
+                       ether_addr_copy(self->mc_list.ar[i++], ha->addr);
 
-               if (i >= AQ_CFG_MULTICAST_ADDRESS_MAX)
-                       break;
+                       if (i >= AQ_HW_MULTICAST_ADDRESS_MAX)
+                               break;
+               }
        }
 
-       if (i >= AQ_CFG_MULTICAST_ADDRESS_MAX) {
-               /* Number of filters is too big: atlantic does not support this.
-                * Force all multi filter to support this.
-                * With this we disable all UC filters and setup "all pass"
-                * multicast mask
-                */
-               self->packet_filter |= IFF_ALLMULTI;
-               self->aq_nic_cfg.mc_list_count = 0;
-               return self->aq_hw_ops->hw_packet_filter_set(self->aq_hw,
-                                                            self->packet_filter);
+       if (i + netdev_mc_count(ndev) > AQ_HW_MULTICAST_ADDRESS_MAX) {
+               packet_filter |= IFF_ALLMULTI;
        } else {
-               return self->aq_hw_ops->hw_multicast_list_set(self->aq_hw,
-                                                   self->mc_list.ar,
-                                                   self->mc_list.count);
+               netdev_for_each_mc_addr(ha, ndev) {
+                       ether_addr_copy(self->mc_list.ar[i++], ha->addr);
+
+                       if (i >= AQ_HW_MULTICAST_ADDRESS_MAX)
+                               break;
+               }
+       }
+
+       if (i > 0 && i < AQ_HW_MULTICAST_ADDRESS_MAX) {
+               packet_filter |= IFF_MULTICAST;
+               self->mc_list.count = i;
+               self->aq_hw_ops->hw_multicast_list_set(self->aq_hw,
+                                                      self->mc_list.ar,
+                                                      self->mc_list.count);
        }
+       return aq_nic_set_packet_filter(self, packet_filter);
 }
 
 int aq_nic_set_mtu(struct aq_nic_s *self, int new_mtu)
index faa533a0ec474116b7d84369c947a7f0f1bfa853..fecfc401f95df041f56f348c1082147377464059 100644 (file)
@@ -75,7 +75,7 @@ struct aq_nic_s {
        struct aq_hw_link_status_s link_status;
        struct {
                u32 count;
-               u8 ar[AQ_CFG_MULTICAST_ADDRESS_MAX][ETH_ALEN];
+               u8 ar[AQ_HW_MULTICAST_ADDRESS_MAX][ETH_ALEN];
        } mc_list;
 
        struct pci_dev *pdev;
index 67e2f9fb9402f3ed419ee46c47a7f6bd4d8e1ffc..8cc6abadc03b90e88fb58b09a53e7da3702710e5 100644 (file)
@@ -765,7 +765,7 @@ static int hw_atl_a0_hw_packet_filter_set(struct aq_hw_s *self,
 
 static int hw_atl_a0_hw_multicast_list_set(struct aq_hw_s *self,
                                           u8 ar_mac
-                                          [AQ_CFG_MULTICAST_ADDRESS_MAX]
+                                          [AQ_HW_MULTICAST_ADDRESS_MAX]
                                           [ETH_ALEN],
                                           u32 count)
 {
index 819f6bcf9b4ee76e620691ae3861a1fad213eca9..956860a697970ab427be0357d8541e929a85c489 100644 (file)
@@ -784,7 +784,7 @@ static int hw_atl_b0_hw_packet_filter_set(struct aq_hw_s *self,
 
 static int hw_atl_b0_hw_multicast_list_set(struct aq_hw_s *self,
                                           u8 ar_mac
-                                          [AQ_CFG_MULTICAST_ADDRESS_MAX]
+                                          [AQ_HW_MULTICAST_ADDRESS_MAX]
                                           [ETH_ALEN],
                                           u32 count)
 {
@@ -812,7 +812,7 @@ static int hw_atl_b0_hw_multicast_list_set(struct aq_hw_s *self,
 
                hw_atl_rpfl2_uc_flr_en_set(self,
                                           (self->aq_nic_cfg->is_mc_list_enabled),
-                                   HW_ATL_B0_MAC_MIN + i);
+                                          HW_ATL_B0_MAC_MIN + i);
        }
 
        err = aq_hw_err_from_flags(self);
index d5fca2e5a9bc34ad6edfa295e378dfe12078c0e5..a1f60f89e05944458e98e7faa2292960368c5ef8 100644 (file)
@@ -1946,8 +1946,8 @@ static int bcm_sysport_open(struct net_device *dev)
        if (!priv->is_lite)
                priv->crc_fwd = !!(umac_readl(priv, UMAC_CMD) & CMD_CRC_FWD);
        else
-               priv->crc_fwd = !!(gib_readl(priv, GIB_CONTROL) &
-                                  GIB_FCS_STRIP);
+               priv->crc_fwd = !((gib_readl(priv, GIB_CONTROL) &
+                                 GIB_FCS_STRIP) >> GIB_FCS_STRIP_SHIFT);
 
        phydev = of_phy_connect(dev, priv->phy_dn, bcm_sysport_adj_link,
                                0, priv->phy_interface);
index d6e5d0cbf3a3b3c526d347add087c9cef428776a..cf440b91fd04a331a7dce529d740a22686b96dfd 100644 (file)
@@ -278,7 +278,8 @@ struct bcm_rsb {
 #define  GIB_GTX_CLK_EXT_CLK           (0 << GIB_GTX_CLK_SEL_SHIFT)
 #define  GIB_GTX_CLK_125MHZ            (1 << GIB_GTX_CLK_SEL_SHIFT)
 #define  GIB_GTX_CLK_250MHZ            (2 << GIB_GTX_CLK_SEL_SHIFT)
-#define  GIB_FCS_STRIP                 (1 << 6)
+#define  GIB_FCS_STRIP_SHIFT           6
+#define  GIB_FCS_STRIP                 (1 << GIB_FCS_STRIP_SHIFT)
 #define  GIB_LCL_LOOP_EN               (1 << 7)
 #define  GIB_LCL_LOOP_TXEN             (1 << 8)
 #define  GIB_RMT_LOOP_EN               (1 << 9)
index 176fc9f4d7defe6a9d5b513902c97f56d732b323..4394c1162be4fde931aa822d69a008b89f54efb8 100644 (file)
@@ -5712,7 +5712,9 @@ static int bnxt_init_chip(struct bnxt *bp, bool irq_re_init)
        }
        vnic->uc_filter_count = 1;
 
-       vnic->rx_mask = CFA_L2_SET_RX_MASK_REQ_MASK_BCAST;
+       vnic->rx_mask = 0;
+       if (bp->dev->flags & IFF_BROADCAST)
+               vnic->rx_mask |= CFA_L2_SET_RX_MASK_REQ_MASK_BCAST;
 
        if ((bp->dev->flags & IFF_PROMISC) && bnxt_promisc_ok(bp))
                vnic->rx_mask |= CFA_L2_SET_RX_MASK_REQ_MASK_PROMISCUOUS;
@@ -5917,7 +5919,7 @@ unsigned int bnxt_get_max_func_irqs(struct bnxt *bp)
        return min_t(unsigned int, hw_resc->max_irqs, hw_resc->max_cp_rings);
 }
 
-void bnxt_set_max_func_irqs(struct bnxt *bp, unsigned int max_irqs)
+static void bnxt_set_max_func_irqs(struct bnxt *bp, unsigned int max_irqs)
 {
        bp->hw_resc.max_irqs = max_irqs;
 }
@@ -6888,7 +6890,7 @@ static int __bnxt_open_nic(struct bnxt *bp, bool irq_re_init, bool link_re_init)
                rc = bnxt_request_irq(bp);
                if (rc) {
                        netdev_err(bp->dev, "bnxt_request_irq err: %x\n", rc);
-                       goto open_err;
+                       goto open_err_irq;
                }
        }
 
@@ -6928,6 +6930,8 @@ static int __bnxt_open_nic(struct bnxt *bp, bool irq_re_init, bool link_re_init)
 open_err:
        bnxt_debug_dev_exit(bp);
        bnxt_disable_napi(bp);
+
+open_err_irq:
        bnxt_del_napi(bp);
 
 open_err_free_mem:
@@ -7214,13 +7218,16 @@ static void bnxt_set_rx_mode(struct net_device *dev)
 
        mask &= ~(CFA_L2_SET_RX_MASK_REQ_MASK_PROMISCUOUS |
                  CFA_L2_SET_RX_MASK_REQ_MASK_MCAST |
-                 CFA_L2_SET_RX_MASK_REQ_MASK_ALL_MCAST);
+                 CFA_L2_SET_RX_MASK_REQ_MASK_ALL_MCAST |
+                 CFA_L2_SET_RX_MASK_REQ_MASK_BCAST);
 
        if ((dev->flags & IFF_PROMISC) && bnxt_promisc_ok(bp))
                mask |= CFA_L2_SET_RX_MASK_REQ_MASK_PROMISCUOUS;
 
        uc_update = bnxt_uc_list_updated(bp);
 
+       if (dev->flags & IFF_BROADCAST)
+               mask |= CFA_L2_SET_RX_MASK_REQ_MASK_BCAST;
        if (dev->flags & IFF_ALLMULTI) {
                mask |= CFA_L2_SET_RX_MASK_REQ_MASK_ALL_MCAST;
                vnic->mc_list_count = 0;
@@ -8502,11 +8509,11 @@ int bnxt_get_max_rings(struct bnxt *bp, int *max_rx, int *max_tx, bool shared)
        int rx, tx, cp;
 
        _bnxt_get_max_rings(bp, &rx, &tx, &cp);
+       *max_rx = rx;
+       *max_tx = tx;
        if (!rx || !tx || !cp)
                return -ENOMEM;
 
-       *max_rx = rx;
-       *max_tx = tx;
        return bnxt_trim_rings(bp, max_rx, max_tx, cp, shared);
 }
 
@@ -8520,8 +8527,11 @@ static int bnxt_get_dflt_rings(struct bnxt *bp, int *max_rx, int *max_tx,
                /* Not enough rings, try disabling agg rings. */
                bp->flags &= ~BNXT_FLAG_AGG_RINGS;
                rc = bnxt_get_max_rings(bp, max_rx, max_tx, shared);
-               if (rc)
+               if (rc) {
+                       /* set BNXT_FLAG_AGG_RINGS back for consistency */
+                       bp->flags |= BNXT_FLAG_AGG_RINGS;
                        return rc;
+               }
                bp->flags |= BNXT_FLAG_NO_AGG_RINGS;
                bp->dev->hw_features &= ~(NETIF_F_LRO | NETIF_F_GRO_HW);
                bp->dev->features &= ~(NETIF_F_LRO | NETIF_F_GRO_HW);
index 9b14eb610b9f653b61092d74b3ab9257a84383d9..91575ef97c8cb119d9407530f4b6f5472d72724c 100644 (file)
@@ -1470,7 +1470,6 @@ void bnxt_set_max_func_stat_ctxs(struct bnxt *bp, unsigned int max);
 unsigned int bnxt_get_max_func_cp_rings(struct bnxt *bp);
 void bnxt_set_max_func_cp_rings(struct bnxt *bp, unsigned int max);
 unsigned int bnxt_get_max_func_irqs(struct bnxt *bp);
-void bnxt_set_max_func_irqs(struct bnxt *bp, unsigned int max);
 int bnxt_get_avail_msix(struct bnxt *bp, int num);
 int bnxt_reserve_rings(struct bnxt *bp);
 void bnxt_tx_disable(struct bnxt *bp);
index 795f45024c209e65591a3e9fe60814315ebb3cb0..491bd40a254d8dad8810d983505b69efe2d011b1 100644 (file)
 #define BNXT_FID_INVALID                       0xffff
 #define VLAN_TCI(vid, prio)    ((vid) | ((prio) << VLAN_PRIO_SHIFT))
 
+#define is_vlan_pcp_wildcarded(vlan_tci_mask)  \
+       ((ntohs(vlan_tci_mask) & VLAN_PRIO_MASK) == 0x0000)
+#define is_vlan_pcp_exactmatch(vlan_tci_mask)  \
+       ((ntohs(vlan_tci_mask) & VLAN_PRIO_MASK) == VLAN_PRIO_MASK)
+#define is_vlan_pcp_zero(vlan_tci)     \
+       ((ntohs(vlan_tci) & VLAN_PRIO_MASK) == 0x0000)
+#define is_vid_exactmatch(vlan_tci_mask)       \
+       ((ntohs(vlan_tci_mask) & VLAN_VID_MASK) == VLAN_VID_MASK)
+
 /* Return the dst fid of the func for flow forwarding
  * For PFs: src_fid is the fid of the PF
  * For VF-reps: src_fid the fid of the VF
@@ -389,6 +398,21 @@ static bool is_exactmatch(void *mask, int len)
        return true;
 }
 
+static bool is_vlan_tci_allowed(__be16  vlan_tci_mask,
+                               __be16  vlan_tci)
+{
+       /* VLAN priority must be either exactly zero or fully wildcarded and
+        * VLAN id must be exact match.
+        */
+       if (is_vid_exactmatch(vlan_tci_mask) &&
+           ((is_vlan_pcp_exactmatch(vlan_tci_mask) &&
+             is_vlan_pcp_zero(vlan_tci)) ||
+            is_vlan_pcp_wildcarded(vlan_tci_mask)))
+               return true;
+
+       return false;
+}
+
 static bool bits_set(void *key, int len)
 {
        const u8 *p = key;
@@ -803,9 +827,9 @@ static bool bnxt_tc_can_offload(struct bnxt *bp, struct bnxt_tc_flow *flow)
        /* Currently VLAN fields cannot be partial wildcard */
        if (bits_set(&flow->l2_key.inner_vlan_tci,
                     sizeof(flow->l2_key.inner_vlan_tci)) &&
-           !is_exactmatch(&flow->l2_mask.inner_vlan_tci,
-                          sizeof(flow->l2_mask.inner_vlan_tci))) {
-               netdev_info(bp->dev, "Wildcard match unsupported for VLAN TCI\n");
+           !is_vlan_tci_allowed(flow->l2_mask.inner_vlan_tci,
+                                flow->l2_key.inner_vlan_tci)) {
+               netdev_info(bp->dev, "Unsupported VLAN TCI\n");
                return false;
        }
        if (bits_set(&flow->l2_key.inner_vlan_tpid,
index 347e4f946eb222ce5c8e1e14777c9d6555eb48dc..840f6e505f733208955bedee497ecf51397d487d 100644 (file)
@@ -169,7 +169,6 @@ static int bnxt_req_msix_vecs(struct bnxt_en_dev *edev, int ulp_id,
                edev->ulp_tbl[ulp_id].msix_requested = avail_msix;
        }
        bnxt_fill_msix_vecs(bp, ent);
-       bnxt_set_max_func_irqs(bp, bnxt_get_max_func_irqs(bp) - avail_msix);
        bnxt_set_max_func_cp_rings(bp, max_cp_rings - avail_msix);
        edev->flags |= BNXT_EN_FLAG_MSIX_REQUESTED;
        return avail_msix;
@@ -192,7 +191,6 @@ static int bnxt_free_msix_vecs(struct bnxt_en_dev *edev, int ulp_id)
        msix_requested = edev->ulp_tbl[ulp_id].msix_requested;
        bnxt_set_max_func_cp_rings(bp, max_cp_rings + msix_requested);
        edev->ulp_tbl[ulp_id].msix_requested = 0;
-       bnxt_set_max_func_irqs(bp, bnxt_get_max_func_irqs(bp) + msix_requested);
        edev->flags &= ~BNXT_EN_FLAG_MSIX_REQUESTED;
        if (netif_running(dev)) {
                bnxt_close_nic(bp, true, false);
index 3be87efdc93d6347da8417ddcd101ed90cc12d8c..aa1374d0af9313dfdbf6a7f8dfeea92e2fee7013 100644 (file)
@@ -6,11 +6,15 @@
  * Copyright (C) 2004 Sun Microsystems Inc.
  * Copyright (C) 2005-2016 Broadcom Corporation.
  * Copyright (C) 2016-2017 Broadcom Limited.
+ * Copyright (C) 2018 Broadcom. All Rights Reserved. The term "Broadcom"
+ * refers to Broadcom Inc. and/or its subsidiaries.
  *
  * Firmware is:
  *     Derived from proprietary unpublished source code,
  *     Copyright (C) 2000-2016 Broadcom Corporation.
  *     Copyright (C) 2016-2017 Broadcom Ltd.
+ *     Copyright (C) 2018 Broadcom. All Rights Reserved. The term "Broadcom"
+ *     refers to Broadcom Inc. and/or its subsidiaries.
  *
  *     Permission is hereby granted for the distribution of this firmware
  *     data in hexadecimal or equivalent format, provided this copyright
@@ -9290,6 +9294,15 @@ static int tg3_chip_reset(struct tg3 *tp)
 
        tg3_restore_clk(tp);
 
+       /* Increase the core clock speed to fix tx timeout issue for 5762
+        * with 100Mbps link speed.
+        */
+       if (tg3_asic_rev(tp) == ASIC_REV_5762) {
+               val = tr32(TG3_CPMU_CLCK_ORIDE_ENABLE);
+               tw32(TG3_CPMU_CLCK_ORIDE_ENABLE, val |
+                    TG3_CPMU_MAC_ORIDE_ENABLE);
+       }
+
        /* Reprobe ASF enable state.  */
        tg3_flag_clear(tp, ENABLE_ASF);
        tp->phy_flags &= ~(TG3_PHYFLG_1G_ON_VAUX_OK |
index 1d61aa3efda177c64c69465f0b72c0df5221ba37..a772a33b685c5eb8c28137107eb33cb4b6ffeb1d 100644 (file)
@@ -7,6 +7,8 @@
  * Copyright (C) 2004 Sun Microsystems Inc.
  * Copyright (C) 2007-2016 Broadcom Corporation.
  * Copyright (C) 2016-2017 Broadcom Limited.
+ * Copyright (C) 2018 Broadcom. All Rights Reserved. The term "Broadcom"
+ * refers to Broadcom Inc. and/or its subsidiaries.
  */
 
 #ifndef _T3_H
index 86659823b2592e20d16e1fbc0640d45c99508f47..3d45f4c92cf6e5d3f091ae654e5312165956d19f 100644 (file)
 #define GEM_DCFG6              0x0294 /* Design Config 6 */
 #define GEM_DCFG7              0x0298 /* Design Config 7 */
 #define GEM_DCFG8              0x029C /* Design Config 8 */
+#define GEM_DCFG10             0x02A4 /* Design Config 10 */
 
 #define GEM_TXBDCTRL   0x04cc /* TX Buffer Descriptor control register */
 #define GEM_RXBDCTRL   0x04d0 /* RX Buffer Descriptor control register */
 #define GEM_SCR2CMP_OFFSET                     0
 #define GEM_SCR2CMP_SIZE                       8
 
+/* Bitfields in DCFG10 */
+#define GEM_TXBD_RDBUFF_OFFSET                 12
+#define GEM_TXBD_RDBUFF_SIZE                   4
+#define GEM_RXBD_RDBUFF_OFFSET                 8
+#define GEM_RXBD_RDBUFF_SIZE                   4
+
 /* Bitfields in TISUBN */
 #define GEM_SUBNSINCR_OFFSET                   0
 #define GEM_SUBNSINCR_SIZE                     16
 #define MACB_CAPS_USRIO_DISABLED               0x00000010
 #define MACB_CAPS_JUMBO                                0x00000020
 #define MACB_CAPS_GEM_HAS_PTP                  0x00000040
+#define MACB_CAPS_BD_RD_PREFETCH               0x00000080
 #define MACB_CAPS_FIFO_MODE                    0x10000000
 #define MACB_CAPS_GIGABIT_MODE_AVAILABLE       0x20000000
 #define MACB_CAPS_SG_DISABLED                  0x40000000
@@ -1203,6 +1211,9 @@ struct macb {
        unsigned int max_tuples;
 
        struct tasklet_struct   hresp_err_tasklet;
+
+       int     rx_bd_rd_prefetch;
+       int     tx_bd_rd_prefetch;
 };
 
 #ifdef CONFIG_MACB_USE_HWSTAMP
index 96cc03a6d9420f52c7c871f52f32c8e9b8973bbb..a6c911bb5ce22588276a9f92561947ff5bff2726 100644 (file)
@@ -1811,23 +1811,25 @@ static void macb_free_consistent(struct macb *bp)
 {
        struct macb_queue *queue;
        unsigned int q;
+       int size;
 
-       queue = &bp->queues[0];
        bp->macbgem_ops.mog_free_rx_buffers(bp);
-       if (queue->rx_ring) {
-               dma_free_coherent(&bp->pdev->dev, RX_RING_BYTES(bp),
-                               queue->rx_ring, queue->rx_ring_dma);
-               queue->rx_ring = NULL;
-       }
 
        for (q = 0, queue = bp->queues; q < bp->num_queues; ++q, ++queue) {
                kfree(queue->tx_skb);
                queue->tx_skb = NULL;
                if (queue->tx_ring) {
-                       dma_free_coherent(&bp->pdev->dev, TX_RING_BYTES(bp),
+                       size = TX_RING_BYTES(bp) + bp->tx_bd_rd_prefetch;
+                       dma_free_coherent(&bp->pdev->dev, size,
                                          queue->tx_ring, queue->tx_ring_dma);
                        queue->tx_ring = NULL;
                }
+               if (queue->rx_ring) {
+                       size = RX_RING_BYTES(bp) + bp->rx_bd_rd_prefetch;
+                       dma_free_coherent(&bp->pdev->dev, size,
+                                         queue->rx_ring, queue->rx_ring_dma);
+                       queue->rx_ring = NULL;
+               }
        }
 }
 
@@ -1874,7 +1876,7 @@ static int macb_alloc_consistent(struct macb *bp)
        int size;
 
        for (q = 0, queue = bp->queues; q < bp->num_queues; ++q, ++queue) {
-               size = TX_RING_BYTES(bp);
+               size = TX_RING_BYTES(bp) + bp->tx_bd_rd_prefetch;
                queue->tx_ring = dma_alloc_coherent(&bp->pdev->dev, size,
                                                    &queue->tx_ring_dma,
                                                    GFP_KERNEL);
@@ -1890,7 +1892,7 @@ static int macb_alloc_consistent(struct macb *bp)
                if (!queue->tx_skb)
                        goto out_err;
 
-               size = RX_RING_BYTES(bp);
+               size = RX_RING_BYTES(bp) + bp->rx_bd_rd_prefetch;
                queue->rx_ring = dma_alloc_coherent(&bp->pdev->dev, size,
                                                 &queue->rx_ring_dma, GFP_KERNEL);
                if (!queue->rx_ring)
@@ -3797,7 +3799,7 @@ static const struct macb_config np4_config = {
 static const struct macb_config zynqmp_config = {
        .caps = MACB_CAPS_GIGABIT_MODE_AVAILABLE |
                        MACB_CAPS_JUMBO |
-                       MACB_CAPS_GEM_HAS_PTP,
+                       MACB_CAPS_GEM_HAS_PTP | MACB_CAPS_BD_RD_PREFETCH,
        .dma_burst_length = 16,
        .clk_init = macb_clk_init,
        .init = macb_init,
@@ -3858,7 +3860,7 @@ static int macb_probe(struct platform_device *pdev)
        void __iomem *mem;
        const char *mac;
        struct macb *bp;
-       int err;
+       int err, val;
 
        regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
        mem = devm_ioremap_resource(&pdev->dev, regs);
@@ -3947,6 +3949,18 @@ static int macb_probe(struct platform_device *pdev)
        else
                dev->max_mtu = ETH_DATA_LEN;
 
+       if (bp->caps & MACB_CAPS_BD_RD_PREFETCH) {
+               val = GEM_BFEXT(RXBD_RDBUFF, gem_readl(bp, DCFG10));
+               if (val)
+                       bp->rx_bd_rd_prefetch = (2 << (val - 1)) *
+                                               macb_dma_desc_get_size(bp);
+
+               val = GEM_BFEXT(TXBD_RDBUFF, gem_readl(bp, DCFG10));
+               if (val)
+                       bp->tx_bd_rd_prefetch = (2 << (val - 1)) *
+                                               macb_dma_desc_get_size(bp);
+       }
+
        mac = of_get_mac_address(np);
        if (mac) {
                ether_addr_copy(bp->dev->dev_addr, mac);
index 043e3c11c42bd407d47561bec2a2e0acd525f12b..92d88c5f76fb8b68e9f8b35ada37d4a77d68f739 100644 (file)
@@ -15,7 +15,7 @@ if NET_VENDOR_CAVIUM
 
 config THUNDER_NIC_PF
        tristate "Thunder Physical function driver"
-       depends on 64BIT
+       depends on 64BIT && PCI
        select THUNDER_NIC_BGX
        ---help---
          This driver supports Thunder's NIC physical function.
@@ -28,13 +28,13 @@ config THUNDER_NIC_PF
 config THUNDER_NIC_VF
        tristate "Thunder Virtual function driver"
        imply CAVIUM_PTP
-       depends on 64BIT
+       depends on 64BIT && PCI
        ---help---
          This driver supports Thunder's NIC virtual function
 
 config THUNDER_NIC_BGX
        tristate "Thunder MAC interface driver (BGX)"
-       depends on 64BIT
+       depends on 64BIT && PCI
        select PHYLIB
        select MDIO_THUNDER
        select THUNDER_NIC_RGX
@@ -44,7 +44,7 @@ config        THUNDER_NIC_BGX
 
 config THUNDER_NIC_RGX
        tristate "Thunder MAC interface driver (RGX)"
-       depends on 64BIT
+       depends on 64BIT && PCI
        select PHYLIB
        select MDIO_THUNDER
        ---help---
@@ -53,7 +53,7 @@ config        THUNDER_NIC_RGX
 
 config CAVIUM_PTP
        tristate "Cavium PTP coprocessor as PTP clock"
-       depends on 64BIT
+       depends on 64BIT && PCI
        imply PTP_1588_CLOCK
        default y
        ---help---
@@ -65,7 +65,7 @@ config CAVIUM_PTP
 
 config LIQUIDIO
        tristate "Cavium LiquidIO support"
-       depends on 64BIT
+       depends on 64BIT && PCI
        depends on MAY_USE_DEVLINK
        imply PTP_1588_CLOCK
        select FW_LOADER
index 8a815bb5717732331293e9fba5b00d3ca23aaf88..7e8454d3b1ad3f382f778c27058695c34b9f13cb 100644 (file)
@@ -91,6 +91,9 @@ static int octeon_console_debug_enabled(u32 console)
  */
 #define LIO_SYNC_OCTEON_TIME_INTERVAL_MS 60000
 
+/* time to wait for possible in-flight requests in milliseconds */
+#define WAIT_INFLIGHT_REQUEST  msecs_to_jiffies(1000)
+
 struct lio_trusted_vf_ctx {
        struct completion complete;
        int status;
@@ -259,7 +262,7 @@ static inline void pcierror_quiesce_device(struct octeon_device *oct)
        force_io_queues_off(oct);
 
        /* To allow for in-flight requests */
-       schedule_timeout_uninterruptible(100);
+       schedule_timeout_uninterruptible(WAIT_INFLIGHT_REQUEST);
 
        if (wait_for_pending_requests(oct))
                dev_err(&oct->pci_dev->dev, "There were pending requests\n");
index 3f6afb54a5eb188061dcad1ce4679465d408db86..bb43ddb7539e719d0cbff780e5ddf17c756dbe05 100644 (file)
@@ -643,13 +643,21 @@ static int octeon_mgmt_set_mac_address(struct net_device *netdev, void *addr)
 static int octeon_mgmt_change_mtu(struct net_device *netdev, int new_mtu)
 {
        struct octeon_mgmt *p = netdev_priv(netdev);
-       int size_without_fcs = new_mtu + OCTEON_MGMT_RX_HEADROOM;
+       int max_packet = new_mtu + ETH_HLEN + ETH_FCS_LEN;
 
        netdev->mtu = new_mtu;
 
-       cvmx_write_csr(p->agl + AGL_GMX_RX_FRM_MAX, size_without_fcs);
+       /* HW lifts the limit if the frame is VLAN tagged
+        * (+4 bytes per each tag, up to two tags)
+        */
+       cvmx_write_csr(p->agl + AGL_GMX_RX_FRM_MAX, max_packet);
+       /* Set the hardware to truncate packets larger than the MTU. The jabber
+        * register must be set to a multiple of 8 bytes, so round up. JABBER is
+        * an unconditional limit, so we need to account for two possible VLAN
+        * tags.
+        */
        cvmx_write_csr(p->agl + AGL_GMX_RX_JABBER,
-                      (size_without_fcs + 7) & 0xfff8);
+                      (max_packet + 7 + VLAN_HLEN * 2) & 0xfff8);
 
        return 0;
 }
index 7b795edd9d3a9543271d29acf0cc35d760a6b065..a19172dbe6be272d9a168302bab18f551a687a17 100644 (file)
@@ -51,6 +51,7 @@
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/uaccess.h>
+#include <linux/nospec.h>
 
 #include "common.h"
 #include "cxgb3_ioctl.h"
@@ -2268,6 +2269,7 @@ static int cxgb_extension_ioctl(struct net_device *dev, void __user *useraddr)
 
                if (t.qset_idx >= nqsets)
                        return -EINVAL;
+               t.qset_idx = array_index_nospec(t.qset_idx, nqsets);
 
                q = &adapter->params.sge.qset[q1 + t.qset_idx];
                t.rspq_size = q->rspq_size;
index 974a868a4824b78dc8cb7225f37b5d2cf8b24b32..3720c3e11ebb883466d04b4a2169878f0b135399 100644 (file)
@@ -8702,7 +8702,7 @@ static int t4_get_flash_params(struct adapter *adap)
        };
 
        unsigned int part, manufacturer;
-       unsigned int density, size;
+       unsigned int density, size = 0;
        u32 flashid = 0;
        int ret;
 
@@ -8772,11 +8772,6 @@ static int t4_get_flash_params(struct adapter *adap)
                case 0x22: /* 256MB */
                        size = 1 << 28;
                        break;
-
-               default:
-                       dev_err(adap->pdev_dev, "Micron Flash Part has bad size, ID = %#x, Density code = %#x\n",
-                               flashid, density);
-                       return -EINVAL;
                }
                break;
        }
@@ -8792,10 +8787,6 @@ static int t4_get_flash_params(struct adapter *adap)
                case 0x17: /* 64MB */
                        size = 1 << 26;
                        break;
-               default:
-                       dev_err(adap->pdev_dev, "ISSI Flash Part has bad size, ID = %#x, Density code = %#x\n",
-                               flashid, density);
-                       return -EINVAL;
                }
                break;
        }
@@ -8811,10 +8802,6 @@ static int t4_get_flash_params(struct adapter *adap)
                case 0x18: /* 16MB */
                        size = 1 << 24;
                        break;
-               default:
-                       dev_err(adap->pdev_dev, "Macronix Flash Part has bad size, ID = %#x, Density code = %#x\n",
-                               flashid, density);
-                       return -EINVAL;
                }
                break;
        }
@@ -8830,17 +8817,21 @@ static int t4_get_flash_params(struct adapter *adap)
                case 0x18: /* 16MB */
                        size = 1 << 24;
                        break;
-               default:
-                       dev_err(adap->pdev_dev, "Winbond Flash Part has bad size, ID = %#x, Density code = %#x\n",
-                               flashid, density);
-                       return -EINVAL;
                }
                break;
        }
-       default:
-               dev_err(adap->pdev_dev, "Unsupported Flash Part, ID = %#x\n",
-                       flashid);
-               return -EINVAL;
+       }
+
+       /* If we didn't recognize the FLASH part, that's no real issue: the
+        * Hardware/Software contract says that Hardware will _*ALWAYS*_
+        * use a FLASH part which is at least 4MB in size and has 64KB
+        * sectors.  The unrecognized FLASH part is likely to be much larger
+        * than 4MB, but that's all we really need.
+        */
+       if (size == 0) {
+               dev_warn(adap->pdev_dev, "Unknown Flash Part, ID = %#x, assuming 4MB\n",
+                        flashid);
+               size = 1 << 22;
        }
 
        /* Store decoded Flash size and fall through into vetting code. */
index d0e196bff0818ce214b4909cb97976aa3502bdc4..ffe7acbeaa22d372b7ce32f9950edb9a98cdf71a 100644 (file)
@@ -329,7 +329,8 @@ static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
        return;
 
 failure:
-       dev_info(dev, "replenish pools failure\n");
+       if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
+               dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
        pool->free_map[pool->next_free] = index;
        pool->rx_buff[index].skb = NULL;
 
@@ -1617,7 +1618,8 @@ static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
                                      &tx_crq);
        }
        if (lpar_rc != H_SUCCESS) {
-               dev_err(dev, "tx failed with code %ld\n", lpar_rc);
+               if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
+                       dev_err_ratelimited(dev, "tx: send failed\n");
                dev_kfree_skb_any(skb);
                tx_buff->skb = NULL;
 
@@ -1825,8 +1827,8 @@ static int do_reset(struct ibmvnic_adapter *adapter,
 
                rc = ibmvnic_login(netdev);
                if (rc) {
-                       adapter->state = VNIC_PROBED;
-                       return 0;
+                       adapter->state = reset_state;
+                       return rc;
                }
 
                if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
@@ -3204,6 +3206,25 @@ static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
        return crq;
 }
 
+static void print_subcrq_error(struct device *dev, int rc, const char *func)
+{
+       switch (rc) {
+       case H_PARAMETER:
+               dev_warn_ratelimited(dev,
+                                    "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
+                                    func, rc);
+               break;
+       case H_CLOSED:
+               dev_warn_ratelimited(dev,
+                                    "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
+                                    func, rc);
+               break;
+       default:
+               dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
+               break;
+       }
+}
+
 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
                       union sub_crq *sub_crq)
 {
@@ -3230,11 +3251,8 @@ static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
                                cpu_to_be64(u64_crq[2]),
                                cpu_to_be64(u64_crq[3]));
 
-       if (rc) {
-               if (rc == H_CLOSED)
-                       dev_warn(dev, "CRQ Queue closed\n");
-               dev_err(dev, "Send error (rc=%d)\n", rc);
-       }
+       if (rc)
+               print_subcrq_error(dev, rc, __func__);
 
        return rc;
 }
@@ -3252,11 +3270,8 @@ static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
                                cpu_to_be64(remote_handle),
                                ioba, num_entries);
 
-       if (rc) {
-               if (rc == H_CLOSED)
-                       dev_warn(dev, "CRQ Queue closed\n");
-               dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
-       }
+       if (rc)
+               print_subcrq_error(dev, rc, __func__);
 
        return rc;
 }
index 3f5c350716bb0e595d79ec928188f5862461694c..0bd1294ba51737240d510f31bbd255faceffeb11 100644 (file)
@@ -1871,7 +1871,12 @@ s32 ixgbe_set_rar_generic(struct ixgbe_hw *hw, u32 index, u8 *addr, u32 vmdq,
        if (enable_addr != 0)
                rar_high |= IXGBE_RAH_AV;
 
+       /* Record lower 32 bits of MAC address and then make
+        * sure that write is flushed to hardware before writing
+        * the upper 16 bits and setting the valid bit.
+        */
        IXGBE_WRITE_REG(hw, IXGBE_RAL(index), rar_low);
+       IXGBE_WRITE_FLUSH(hw);
        IXGBE_WRITE_REG(hw, IXGBE_RAH(index), rar_high);
 
        return 0;
@@ -1903,8 +1908,13 @@ s32 ixgbe_clear_rar_generic(struct ixgbe_hw *hw, u32 index)
        rar_high = IXGBE_READ_REG(hw, IXGBE_RAH(index));
        rar_high &= ~(0x0000FFFF | IXGBE_RAH_AV);
 
-       IXGBE_WRITE_REG(hw, IXGBE_RAL(index), 0);
+       /* Clear the address valid bit and upper 16 bits of the address
+        * before clearing the lower bits. This way we aren't updating
+        * a live filter.
+        */
        IXGBE_WRITE_REG(hw, IXGBE_RAH(index), rar_high);
+       IXGBE_WRITE_FLUSH(hw);
+       IXGBE_WRITE_REG(hw, IXGBE_RAL(index), 0);
 
        /* clear VMDq pool/queue selection for this RAR */
        hw->mac.ops.clear_vmdq(hw, index, IXGBE_CLEAR_VMDQ_ALL);
index c116f459945d62455843d4e9262971630dd45099..da4322e4daed5de4fb44f06d8cdb488bc41f6432 100644 (file)
@@ -839,7 +839,7 @@ int ixgbe_ipsec_tx(struct ixgbe_ring *tx_ring,
        }
 
        itd->sa_idx = xs->xso.offload_handle - IXGBE_IPSEC_BASE_TX_INDEX;
-       if (unlikely(itd->sa_idx > IXGBE_IPSEC_MAX_SA_COUNT)) {
+       if (unlikely(itd->sa_idx >= IXGBE_IPSEC_MAX_SA_COUNT)) {
                netdev_err(tx_ring->netdev, "%s: bad sa_idx=%d handle=%lu\n",
                           __func__, itd->sa_idx, xs->xso.offload_handle);
                return 0;
index 9f54ccbddea74b57973ee724acf360fa23434a3e..3360f7b9ee73bdb32957472299a3438e8189f7bd 100644 (file)
@@ -474,10 +474,10 @@ static int mlx4_en_complete_rx_desc(struct mlx4_en_priv *priv,
 {
        const struct mlx4_en_frag_info *frag_info = priv->frag_info;
        unsigned int truesize = 0;
+       bool release = true;
        int nr, frag_size;
        struct page *page;
        dma_addr_t dma;
-       bool release;
 
        /* Collect used fragments while replacing them in the HW descriptors */
        for (nr = 0;; frags++) {
@@ -500,7 +500,11 @@ static int mlx4_en_complete_rx_desc(struct mlx4_en_priv *priv,
                        release = page_count(page) != 1 ||
                                  page_is_pfmemalloc(page) ||
                                  page_to_nid(page) != numa_mem_id();
-               } else {
+               } else if (!priv->rx_headroom) {
+                       /* rx_headroom for non XDP setup is always 0.
+                        * When XDP is set, the above condition will
+                        * guarantee page is always released.
+                        */
                        u32 sz_align = ALIGN(frag_size, SMP_CACHE_BYTES);
 
                        frags->page_offset += sz_align;
index 6aaaf3d9ba31d9538d9307caa0450a848bf6b091..77b2adb293415a9de16caaabbd203b397cd12a4a 100644 (file)
@@ -4756,6 +4756,12 @@ static void mlxsw_sp_rt6_destroy(struct mlxsw_sp_rt6 *mlxsw_sp_rt6)
        kfree(mlxsw_sp_rt6);
 }
 
+static bool mlxsw_sp_fib6_rt_can_mp(const struct fib6_info *rt)
+{
+       /* RTF_CACHE routes are ignored */
+       return (rt->fib6_flags & (RTF_GATEWAY | RTF_ADDRCONF)) == RTF_GATEWAY;
+}
+
 static struct fib6_info *
 mlxsw_sp_fib6_entry_rt(const struct mlxsw_sp_fib6_entry *fib6_entry)
 {
@@ -4765,11 +4771,11 @@ mlxsw_sp_fib6_entry_rt(const struct mlxsw_sp_fib6_entry *fib6_entry)
 
 static struct mlxsw_sp_fib6_entry *
 mlxsw_sp_fib6_node_mp_entry_find(const struct mlxsw_sp_fib_node *fib_node,
-                                const struct fib6_info *nrt, bool append)
+                                const struct fib6_info *nrt, bool replace)
 {
        struct mlxsw_sp_fib6_entry *fib6_entry;
 
-       if (!append)
+       if (!mlxsw_sp_fib6_rt_can_mp(nrt) || replace)
                return NULL;
 
        list_for_each_entry(fib6_entry, &fib_node->entry_list, common.list) {
@@ -4784,7 +4790,8 @@ mlxsw_sp_fib6_node_mp_entry_find(const struct mlxsw_sp_fib_node *fib_node,
                        break;
                if (rt->fib6_metric < nrt->fib6_metric)
                        continue;
-               if (rt->fib6_metric == nrt->fib6_metric)
+               if (rt->fib6_metric == nrt->fib6_metric &&
+                   mlxsw_sp_fib6_rt_can_mp(rt))
                        return fib6_entry;
                if (rt->fib6_metric > nrt->fib6_metric)
                        break;
@@ -5163,7 +5170,7 @@ static struct mlxsw_sp_fib6_entry *
 mlxsw_sp_fib6_node_entry_find(const struct mlxsw_sp_fib_node *fib_node,
                              const struct fib6_info *nrt, bool replace)
 {
-       struct mlxsw_sp_fib6_entry *fib6_entry;
+       struct mlxsw_sp_fib6_entry *fib6_entry, *fallback = NULL;
 
        list_for_each_entry(fib6_entry, &fib_node->entry_list, common.list) {
                struct fib6_info *rt = mlxsw_sp_fib6_entry_rt(fib6_entry);
@@ -5172,13 +5179,18 @@ mlxsw_sp_fib6_node_entry_find(const struct mlxsw_sp_fib_node *fib_node,
                        continue;
                if (rt->fib6_table->tb6_id != nrt->fib6_table->tb6_id)
                        break;
-               if (replace && rt->fib6_metric == nrt->fib6_metric)
-                       return fib6_entry;
+               if (replace && rt->fib6_metric == nrt->fib6_metric) {
+                       if (mlxsw_sp_fib6_rt_can_mp(rt) ==
+                           mlxsw_sp_fib6_rt_can_mp(nrt))
+                               return fib6_entry;
+                       if (mlxsw_sp_fib6_rt_can_mp(nrt))
+                               fallback = fallback ?: fib6_entry;
+               }
                if (rt->fib6_metric > nrt->fib6_metric)
-                       return fib6_entry;
+                       return fallback ?: fib6_entry;
        }
 
-       return NULL;
+       return fallback;
 }
 
 static int
@@ -5304,8 +5316,7 @@ static void mlxsw_sp_fib6_entry_replace(struct mlxsw_sp *mlxsw_sp,
 }
 
 static int mlxsw_sp_router_fib6_add(struct mlxsw_sp *mlxsw_sp,
-                                   struct fib6_info *rt, bool replace,
-                                   bool append)
+                                   struct fib6_info *rt, bool replace)
 {
        struct mlxsw_sp_fib6_entry *fib6_entry;
        struct mlxsw_sp_fib_node *fib_node;
@@ -5331,7 +5342,7 @@ static int mlxsw_sp_router_fib6_add(struct mlxsw_sp *mlxsw_sp,
        /* Before creating a new entry, try to append route to an existing
         * multipath entry.
         */
-       fib6_entry = mlxsw_sp_fib6_node_mp_entry_find(fib_node, rt, append);
+       fib6_entry = mlxsw_sp_fib6_node_mp_entry_find(fib_node, rt, replace);
        if (fib6_entry) {
                err = mlxsw_sp_fib6_entry_nexthop_add(mlxsw_sp, fib6_entry, rt);
                if (err)
@@ -5339,14 +5350,6 @@ static int mlxsw_sp_router_fib6_add(struct mlxsw_sp *mlxsw_sp,
                return 0;
        }
 
-       /* We received an append event, yet did not find any route to
-        * append to.
-        */
-       if (WARN_ON(append)) {
-               err = -EINVAL;
-               goto err_fib6_entry_append;
-       }
-
        fib6_entry = mlxsw_sp_fib6_entry_create(mlxsw_sp, fib_node, rt);
        if (IS_ERR(fib6_entry)) {
                err = PTR_ERR(fib6_entry);
@@ -5364,7 +5367,6 @@ static int mlxsw_sp_router_fib6_add(struct mlxsw_sp *mlxsw_sp,
 err_fib6_node_entry_link:
        mlxsw_sp_fib6_entry_destroy(mlxsw_sp, fib6_entry);
 err_fib6_entry_create:
-err_fib6_entry_append:
 err_fib6_entry_nexthop_add:
        mlxsw_sp_fib_node_put(mlxsw_sp, fib_node);
        return err;
@@ -5715,7 +5717,7 @@ static void mlxsw_sp_router_fib6_event_work(struct work_struct *work)
        struct mlxsw_sp_fib_event_work *fib_work =
                container_of(work, struct mlxsw_sp_fib_event_work, work);
        struct mlxsw_sp *mlxsw_sp = fib_work->mlxsw_sp;
-       bool replace, append;
+       bool replace;
        int err;
 
        rtnl_lock();
@@ -5726,10 +5728,8 @@ static void mlxsw_sp_router_fib6_event_work(struct work_struct *work)
        case FIB_EVENT_ENTRY_APPEND: /* fall through */
        case FIB_EVENT_ENTRY_ADD:
                replace = fib_work->event == FIB_EVENT_ENTRY_REPLACE;
-               append = fib_work->event == FIB_EVENT_ENTRY_APPEND;
                err = mlxsw_sp_router_fib6_add(mlxsw_sp,
-                                              fib_work->fen6_info.rt, replace,
-                                              append);
+                                              fib_work->fen6_info.rt, replace);
                if (err)
                        mlxsw_sp_router_fib_abort(mlxsw_sp);
                mlxsw_sp_rt6_release(fib_work->fen6_info.rt);
index 00db3401b89852a7fe5eaca7342344bcb3b66d4d..1dfaccd151f0d457a2ce38447400925113ae546d 100644 (file)
@@ -502,6 +502,7 @@ enum BAR_ID {
 struct qed_nvm_image_info {
        u32 num_images;
        struct bist_nvm_image_att *image_att;
+       bool valid;
 };
 
 #define DRV_MODULE_VERSION                   \
index a14e484890299565ee8fdac8851ed9d7f3e90437..4340c4c90bcbe8b03e5373cfc674c8840ff640d9 100644 (file)
@@ -6723,7 +6723,7 @@ static enum dbg_status qed_parse_mcp_trace_buf(u8 *trace_buf,
                format_idx = header & MFW_TRACE_EVENTID_MASK;
 
                /* Skip message if its index doesn't exist in the meta data */
-               if (format_idx > s_mcp_trace_meta.formats_num) {
+               if (format_idx >= s_mcp_trace_meta.formats_num) {
                        u8 format_size =
                                (u8)((header & MFW_TRACE_PRM_SIZE_MASK) >>
                                     MFW_TRACE_PRM_SIZE_SHIFT);
index 0cbc74d6ca8b4914b01c4e12a253a59dcee21bf8..758a9a5127fa8c00566e4f90d5f75db636570e33 100644 (file)
@@ -371,7 +371,7 @@ static struct qed_dev *qed_probe(struct pci_dev *pdev,
                goto err2;
        }
 
-       DP_INFO(cdev, "qed_probe completed successffuly\n");
+       DP_INFO(cdev, "qed_probe completed successfully\n");
 
        return cdev;
 
index 4e0b443c9519d67bc3b888ddf3b341c93291e328..9d9e533bccdcd351452753fbb0600db098685eb9 100644 (file)
@@ -592,6 +592,9 @@ int qed_mcp_nvm_wr_cmd(struct qed_hwfn *p_hwfn,
        *o_mcp_resp = mb_params.mcp_resp;
        *o_mcp_param = mb_params.mcp_param;
 
+       /* nvm_info needs to be updated */
+       p_hwfn->nvm_info.valid = false;
+
        return 0;
 }
 
@@ -2555,11 +2558,14 @@ int qed_mcp_bist_nvm_get_image_att(struct qed_hwfn *p_hwfn,
 
 int qed_mcp_nvm_info_populate(struct qed_hwfn *p_hwfn)
 {
-       struct qed_nvm_image_info *nvm_info = &p_hwfn->nvm_info;
+       struct qed_nvm_image_info nvm_info;
        struct qed_ptt *p_ptt;
        int rc;
        u32 i;
 
+       if (p_hwfn->nvm_info.valid)
+               return 0;
+
        p_ptt = qed_ptt_acquire(p_hwfn);
        if (!p_ptt) {
                DP_ERR(p_hwfn, "failed to acquire ptt\n");
@@ -2567,29 +2573,29 @@ int qed_mcp_nvm_info_populate(struct qed_hwfn *p_hwfn)
        }
 
        /* Acquire from MFW the amount of available images */
-       nvm_info->num_images = 0;
+       nvm_info.num_images = 0;
        rc = qed_mcp_bist_nvm_get_num_images(p_hwfn,
-                                            p_ptt, &nvm_info->num_images);
+                                            p_ptt, &nvm_info.num_images);
        if (rc == -EOPNOTSUPP) {
                DP_INFO(p_hwfn, "DRV_MSG_CODE_BIST_TEST is not supported\n");
                goto out;
-       } else if (rc || !nvm_info->num_images) {
+       } else if (rc || !nvm_info.num_images) {
                DP_ERR(p_hwfn, "Failed getting number of images\n");
                goto err0;
        }
 
-       nvm_info->image_att = kmalloc_array(nvm_info->num_images,
-                                           sizeof(struct bist_nvm_image_att),
-                                           GFP_KERNEL);
-       if (!nvm_info->image_att) {
+       nvm_info.image_att = kmalloc_array(nvm_info.num_images,
+                                          sizeof(struct bist_nvm_image_att),
+                                          GFP_KERNEL);
+       if (!nvm_info.image_att) {
                rc = -ENOMEM;
                goto err0;
        }
 
        /* Iterate over images and get their attributes */
-       for (i = 0; i < nvm_info->num_images; i++) {
+       for (i = 0; i < nvm_info.num_images; i++) {
                rc = qed_mcp_bist_nvm_get_image_att(p_hwfn, p_ptt,
-                                                   &nvm_info->image_att[i], i);
+                                                   &nvm_info.image_att[i], i);
                if (rc) {
                        DP_ERR(p_hwfn,
                               "Failed getting image index %d attributes\n", i);
@@ -2597,14 +2603,22 @@ int qed_mcp_nvm_info_populate(struct qed_hwfn *p_hwfn)
                }
 
                DP_VERBOSE(p_hwfn, QED_MSG_SP, "image index %d, size %x\n", i,
-                          nvm_info->image_att[i].len);
+                          nvm_info.image_att[i].len);
        }
 out:
+       /* Update hwfn's nvm_info */
+       if (nvm_info.num_images) {
+               p_hwfn->nvm_info.num_images = nvm_info.num_images;
+               kfree(p_hwfn->nvm_info.image_att);
+               p_hwfn->nvm_info.image_att = nvm_info.image_att;
+               p_hwfn->nvm_info.valid = true;
+       }
+
        qed_ptt_release(p_hwfn, p_ptt);
        return 0;
 
 err1:
-       kfree(nvm_info->image_att);
+       kfree(nvm_info.image_att);
 err0:
        qed_ptt_release(p_hwfn, p_ptt);
        return rc;
@@ -2641,6 +2655,7 @@ qed_mcp_get_nvm_image_att(struct qed_hwfn *p_hwfn,
                return -EINVAL;
        }
 
+       qed_mcp_nvm_info_populate(p_hwfn);
        for (i = 0; i < p_hwfn->nvm_info.num_images; i++)
                if (type == p_hwfn->nvm_info.image_att[i].image_type)
                        break;
index 891f03a7a33dc7286b5bb6d1b4ac2333ab74aacf..8d7b9bb910f2addae4712088884b334c42876934 100644 (file)
@@ -1128,6 +1128,8 @@ static ssize_t qlcnic_83xx_sysfs_flash_write_handler(struct file *filp,
        struct qlcnic_adapter *adapter = dev_get_drvdata(dev);
 
        ret = kstrtoul(buf, 16, &data);
+       if (ret)
+               return ret;
 
        switch (data) {
        case QLC_83XX_FLASH_SECTOR_ERASE_CMD:
index 5803cd6db406c7f9c5426ceb87bf062d4f0434fb..206f0266463e362a0e34fe8ff5b626519500e2ed 100644 (file)
@@ -658,7 +658,7 @@ qcaspi_netdev_open(struct net_device *dev)
                return ret;
        }
 
-       netif_start_queue(qca->net_dev);
+       /* SPI thread takes care of TX queue */
 
        return 0;
 }
@@ -760,6 +760,9 @@ qcaspi_netdev_tx_timeout(struct net_device *dev)
        qca->net_dev->stats.tx_errors++;
        /* Trigger tx queue flush and QCA7000 reset */
        qca->sync = QCASPI_SYNC_UNKNOWN;
+
+       if (qca->spi_thread)
+               wake_up_process(qca->spi_thread);
 }
 
 static int
@@ -878,22 +881,22 @@ qca_spi_probe(struct spi_device *spi)
 
        if ((qcaspi_clkspeed < QCASPI_CLK_SPEED_MIN) ||
            (qcaspi_clkspeed > QCASPI_CLK_SPEED_MAX)) {
-               dev_info(&spi->dev, "Invalid clkspeed: %d\n",
-                        qcaspi_clkspeed);
+               dev_err(&spi->dev, "Invalid clkspeed: %d\n",
+                       qcaspi_clkspeed);
                return -EINVAL;
        }
 
        if ((qcaspi_burst_len < QCASPI_BURST_LEN_MIN) ||
            (qcaspi_burst_len > QCASPI_BURST_LEN_MAX)) {
-               dev_info(&spi->dev, "Invalid burst len: %d\n",
-                        qcaspi_burst_len);
+               dev_err(&spi->dev, "Invalid burst len: %d\n",
+                       qcaspi_burst_len);
                return -EINVAL;
        }
 
        if ((qcaspi_pluggable < QCASPI_PLUGGABLE_MIN) ||
            (qcaspi_pluggable > QCASPI_PLUGGABLE_MAX)) {
-               dev_info(&spi->dev, "Invalid pluggable: %d\n",
-                        qcaspi_pluggable);
+               dev_err(&spi->dev, "Invalid pluggable: %d\n",
+                       qcaspi_pluggable);
                return -EINVAL;
        }
 
@@ -955,8 +958,8 @@ qca_spi_probe(struct spi_device *spi)
        }
 
        if (register_netdev(qcaspi_devs)) {
-               dev_info(&spi->dev, "Unable to register net device %s\n",
-                        qcaspi_devs->name);
+               dev_err(&spi->dev, "Unable to register net device %s\n",
+                       qcaspi_devs->name);
                free_netdev(qcaspi_devs);
                return -EFAULT;
        }
index f4cae2be0fda563e71f5b07f1798655169c8b683..a3f69901ac877a058fd477d1ef7d3aac984eeded 100644 (file)
@@ -7789,6 +7789,7 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
                NETIF_F_HW_VLAN_CTAG_RX;
        dev->vlan_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_TSO |
                NETIF_F_HIGHDMA;
+       dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
 
        tp->cp_cmd |= RxChkSum | RxVlan;
 
index 68f122140966d4de381b47fa192246eb7606707a..0d811c02ff340f09a385ec0677f0388034615eef 100644 (file)
@@ -980,6 +980,13 @@ static void ravb_adjust_link(struct net_device *ndev)
        struct ravb_private *priv = netdev_priv(ndev);
        struct phy_device *phydev = ndev->phydev;
        bool new_state = false;
+       unsigned long flags;
+
+       spin_lock_irqsave(&priv->lock, flags);
+
+       /* Disable TX and RX right over here, if E-MAC change is ignored */
+       if (priv->no_avb_link)
+               ravb_rcv_snd_disable(ndev);
 
        if (phydev->link) {
                if (phydev->duplex != priv->duplex) {
@@ -997,18 +1004,21 @@ static void ravb_adjust_link(struct net_device *ndev)
                        ravb_modify(ndev, ECMR, ECMR_TXF, 0);
                        new_state = true;
                        priv->link = phydev->link;
-                       if (priv->no_avb_link)
-                               ravb_rcv_snd_enable(ndev);
                }
        } else if (priv->link) {
                new_state = true;
                priv->link = 0;
                priv->speed = 0;
                priv->duplex = -1;
-               if (priv->no_avb_link)
-                       ravb_rcv_snd_disable(ndev);
        }
 
+       /* Enable TX and RX right over here, if E-MAC change is ignored */
+       if (priv->no_avb_link && phydev->link)
+               ravb_rcv_snd_enable(ndev);
+
+       mmiowb();
+       spin_unlock_irqrestore(&priv->lock, flags);
+
        if (new_state && netif_msg_link(priv))
                phy_print_status(phydev);
 }
@@ -1096,75 +1106,6 @@ static int ravb_phy_start(struct net_device *ndev)
        return 0;
 }
 
-static int ravb_get_link_ksettings(struct net_device *ndev,
-                                  struct ethtool_link_ksettings *cmd)
-{
-       struct ravb_private *priv = netdev_priv(ndev);
-       unsigned long flags;
-
-       if (!ndev->phydev)
-               return -ENODEV;
-
-       spin_lock_irqsave(&priv->lock, flags);
-       phy_ethtool_ksettings_get(ndev->phydev, cmd);
-       spin_unlock_irqrestore(&priv->lock, flags);
-
-       return 0;
-}
-
-static int ravb_set_link_ksettings(struct net_device *ndev,
-                                  const struct ethtool_link_ksettings *cmd)
-{
-       struct ravb_private *priv = netdev_priv(ndev);
-       unsigned long flags;
-       int error;
-
-       if (!ndev->phydev)
-               return -ENODEV;
-
-       spin_lock_irqsave(&priv->lock, flags);
-
-       /* Disable TX and RX */
-       ravb_rcv_snd_disable(ndev);
-
-       error = phy_ethtool_ksettings_set(ndev->phydev, cmd);
-       if (error)
-               goto error_exit;
-
-       if (cmd->base.duplex == DUPLEX_FULL)
-               priv->duplex = 1;
-       else
-               priv->duplex = 0;
-
-       ravb_set_duplex(ndev);
-
-error_exit:
-       mdelay(1);
-
-       /* Enable TX and RX */
-       ravb_rcv_snd_enable(ndev);
-
-       mmiowb();
-       spin_unlock_irqrestore(&priv->lock, flags);
-
-       return error;
-}
-
-static int ravb_nway_reset(struct net_device *ndev)
-{
-       struct ravb_private *priv = netdev_priv(ndev);
-       int error = -ENODEV;
-       unsigned long flags;
-
-       if (ndev->phydev) {
-               spin_lock_irqsave(&priv->lock, flags);
-               error = phy_start_aneg(ndev->phydev);
-               spin_unlock_irqrestore(&priv->lock, flags);
-       }
-
-       return error;
-}
-
 static u32 ravb_get_msglevel(struct net_device *ndev)
 {
        struct ravb_private *priv = netdev_priv(ndev);
@@ -1377,7 +1318,7 @@ static int ravb_set_wol(struct net_device *ndev, struct ethtool_wolinfo *wol)
 }
 
 static const struct ethtool_ops ravb_ethtool_ops = {
-       .nway_reset             = ravb_nway_reset,
+       .nway_reset             = phy_ethtool_nway_reset,
        .get_msglevel           = ravb_get_msglevel,
        .set_msglevel           = ravb_set_msglevel,
        .get_link               = ethtool_op_get_link,
@@ -1387,8 +1328,8 @@ static const struct ethtool_ops ravb_ethtool_ops = {
        .get_ringparam          = ravb_get_ringparam,
        .set_ringparam          = ravb_set_ringparam,
        .get_ts_info            = ravb_get_ts_info,
-       .get_link_ksettings     = ravb_get_link_ksettings,
-       .set_link_ksettings     = ravb_set_link_ksettings,
+       .get_link_ksettings     = phy_ethtool_get_link_ksettings,
+       .set_link_ksettings     = phy_ethtool_set_link_ksettings,
        .get_wol                = ravb_get_wol,
        .set_wol                = ravb_set_wol,
 };
index e9007b613f17ca8de16b67e054df42a800522fb5..5614fd231bbe1e4685582e15faf27dad412b241b 100644 (file)
@@ -1927,8 +1927,15 @@ static void sh_eth_adjust_link(struct net_device *ndev)
 {
        struct sh_eth_private *mdp = netdev_priv(ndev);
        struct phy_device *phydev = ndev->phydev;
+       unsigned long flags;
        int new_state = 0;
 
+       spin_lock_irqsave(&mdp->lock, flags);
+
+       /* Disable TX and RX right over here, if E-MAC change is ignored */
+       if (mdp->cd->no_psr || mdp->no_ether_link)
+               sh_eth_rcv_snd_disable(ndev);
+
        if (phydev->link) {
                if (phydev->duplex != mdp->duplex) {
                        new_state = 1;
@@ -1947,18 +1954,21 @@ static void sh_eth_adjust_link(struct net_device *ndev)
                        sh_eth_modify(ndev, ECMR, ECMR_TXF, 0);
                        new_state = 1;
                        mdp->link = phydev->link;
-                       if (mdp->cd->no_psr || mdp->no_ether_link)
-                               sh_eth_rcv_snd_enable(ndev);
                }
        } else if (mdp->link) {
                new_state = 1;
                mdp->link = 0;
                mdp->speed = 0;
                mdp->duplex = -1;
-               if (mdp->cd->no_psr || mdp->no_ether_link)
-                       sh_eth_rcv_snd_disable(ndev);
        }
 
+       /* Enable TX and RX right over here, if E-MAC change is ignored */
+       if ((mdp->cd->no_psr || mdp->no_ether_link) && phydev->link)
+               sh_eth_rcv_snd_enable(ndev);
+
+       mmiowb();
+       spin_unlock_irqrestore(&mdp->lock, flags);
+
        if (new_state && netif_msg_link(mdp))
                phy_print_status(phydev);
 }
@@ -2030,60 +2040,6 @@ static int sh_eth_phy_start(struct net_device *ndev)
        return 0;
 }
 
-static int sh_eth_get_link_ksettings(struct net_device *ndev,
-                                    struct ethtool_link_ksettings *cmd)
-{
-       struct sh_eth_private *mdp = netdev_priv(ndev);
-       unsigned long flags;
-
-       if (!ndev->phydev)
-               return -ENODEV;
-
-       spin_lock_irqsave(&mdp->lock, flags);
-       phy_ethtool_ksettings_get(ndev->phydev, cmd);
-       spin_unlock_irqrestore(&mdp->lock, flags);
-
-       return 0;
-}
-
-static int sh_eth_set_link_ksettings(struct net_device *ndev,
-                                    const struct ethtool_link_ksettings *cmd)
-{
-       struct sh_eth_private *mdp = netdev_priv(ndev);
-       unsigned long flags;
-       int ret;
-
-       if (!ndev->phydev)
-               return -ENODEV;
-
-       spin_lock_irqsave(&mdp->lock, flags);
-
-       /* disable tx and rx */
-       sh_eth_rcv_snd_disable(ndev);
-
-       ret = phy_ethtool_ksettings_set(ndev->phydev, cmd);
-       if (ret)
-               goto error_exit;
-
-       if (cmd->base.duplex == DUPLEX_FULL)
-               mdp->duplex = 1;
-       else
-               mdp->duplex = 0;
-
-       if (mdp->cd->set_duplex)
-               mdp->cd->set_duplex(ndev);
-
-error_exit:
-       mdelay(1);
-
-       /* enable tx and rx */
-       sh_eth_rcv_snd_enable(ndev);
-
-       spin_unlock_irqrestore(&mdp->lock, flags);
-
-       return ret;
-}
-
 /* If it is ever necessary to increase SH_ETH_REG_DUMP_MAX_REGS, the
  * version must be bumped as well.  Just adding registers up to that
  * limit is fine, as long as the existing register indices don't
@@ -2263,22 +2219,6 @@ static void sh_eth_get_regs(struct net_device *ndev, struct ethtool_regs *regs,
        pm_runtime_put_sync(&mdp->pdev->dev);
 }
 
-static int sh_eth_nway_reset(struct net_device *ndev)
-{
-       struct sh_eth_private *mdp = netdev_priv(ndev);
-       unsigned long flags;
-       int ret;
-
-       if (!ndev->phydev)
-               return -ENODEV;
-
-       spin_lock_irqsave(&mdp->lock, flags);
-       ret = phy_start_aneg(ndev->phydev);
-       spin_unlock_irqrestore(&mdp->lock, flags);
-
-       return ret;
-}
-
 static u32 sh_eth_get_msglevel(struct net_device *ndev)
 {
        struct sh_eth_private *mdp = netdev_priv(ndev);
@@ -2429,7 +2369,7 @@ static int sh_eth_set_wol(struct net_device *ndev, struct ethtool_wolinfo *wol)
 static const struct ethtool_ops sh_eth_ethtool_ops = {
        .get_regs_len   = sh_eth_get_regs_len,
        .get_regs       = sh_eth_get_regs,
-       .nway_reset     = sh_eth_nway_reset,
+       .nway_reset     = phy_ethtool_nway_reset,
        .get_msglevel   = sh_eth_get_msglevel,
        .set_msglevel   = sh_eth_set_msglevel,
        .get_link       = ethtool_op_get_link,
@@ -2438,8 +2378,8 @@ static const struct ethtool_ops sh_eth_ethtool_ops = {
        .get_sset_count     = sh_eth_get_sset_count,
        .get_ringparam  = sh_eth_get_ringparam,
        .set_ringparam  = sh_eth_set_ringparam,
-       .get_link_ksettings = sh_eth_get_link_ksettings,
-       .set_link_ksettings = sh_eth_set_link_ksettings,
+       .get_link_ksettings = phy_ethtool_get_link_ksettings,
+       .set_link_ksettings = phy_ethtool_set_link_ksettings,
        .get_wol        = sh_eth_get_wol,
        .set_wol        = sh_eth_set_wol,
 };
index 23f0785c0573ec72fea3db10dfdf353c41341ee8..7eeac3d6cfe898a9a4ef6df9378d8c6d29383ce1 100644 (file)
@@ -4288,9 +4288,9 @@ static int efx_ef10_filter_pri(struct efx_ef10_filter_table *table,
        return -EPROTONOSUPPORT;
 }
 
-static s32 efx_ef10_filter_insert(struct efx_nic *efx,
-                                 struct efx_filter_spec *spec,
-                                 bool replace_equal)
+static s32 efx_ef10_filter_insert_locked(struct efx_nic *efx,
+                                        struct efx_filter_spec *spec,
+                                        bool replace_equal)
 {
        DECLARE_BITMAP(mc_rem_map, EFX_EF10_FILTER_SEARCH_LIMIT);
        struct efx_ef10_nic_data *nic_data = efx->nic_data;
@@ -4307,7 +4307,7 @@ static s32 efx_ef10_filter_insert(struct efx_nic *efx,
        bool is_mc_recip;
        s32 rc;
 
-       down_read(&efx->filter_sem);
+       WARN_ON(!rwsem_is_locked(&efx->filter_sem));
        table = efx->filter_state;
        down_write(&table->lock);
 
@@ -4498,10 +4498,22 @@ out_unlock:
        if (rss_locked)
                mutex_unlock(&efx->rss_lock);
        up_write(&table->lock);
-       up_read(&efx->filter_sem);
        return rc;
 }
 
+static s32 efx_ef10_filter_insert(struct efx_nic *efx,
+                                 struct efx_filter_spec *spec,
+                                 bool replace_equal)
+{
+       s32 ret;
+
+       down_read(&efx->filter_sem);
+       ret = efx_ef10_filter_insert_locked(efx, spec, replace_equal);
+       up_read(&efx->filter_sem);
+
+       return ret;
+}
+
 static void efx_ef10_filter_update_rx_scatter(struct efx_nic *efx)
 {
        /* no need to do anything here on EF10 */
@@ -5285,7 +5297,7 @@ static int efx_ef10_filter_insert_addr_list(struct efx_nic *efx,
                EFX_WARN_ON_PARANOID(ids[i] != EFX_EF10_FILTER_ID_INVALID);
                efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO, filter_flags, 0);
                efx_filter_set_eth_local(&spec, vlan->vid, addr_list[i].addr);
-               rc = efx_ef10_filter_insert(efx, &spec, true);
+               rc = efx_ef10_filter_insert_locked(efx, &spec, true);
                if (rc < 0) {
                        if (rollback) {
                                netif_info(efx, drv, efx->net_dev,
@@ -5314,7 +5326,7 @@ static int efx_ef10_filter_insert_addr_list(struct efx_nic *efx,
                efx_filter_init_rx(&spec, EFX_FILTER_PRI_AUTO, filter_flags, 0);
                eth_broadcast_addr(baddr);
                efx_filter_set_eth_local(&spec, vlan->vid, baddr);
-               rc = efx_ef10_filter_insert(efx, &spec, true);
+               rc = efx_ef10_filter_insert_locked(efx, &spec, true);
                if (rc < 0) {
                        netif_warn(efx, drv, efx->net_dev,
                                   "Broadcast filter insert failed rc=%d\n", rc);
@@ -5370,7 +5382,7 @@ static int efx_ef10_filter_insert_def(struct efx_nic *efx,
        if (vlan->vid != EFX_FILTER_VID_UNSPEC)
                efx_filter_set_eth_local(&spec, vlan->vid, NULL);
 
-       rc = efx_ef10_filter_insert(efx, &spec, true);
+       rc = efx_ef10_filter_insert_locked(efx, &spec, true);
        if (rc < 0) {
                const char *um = multicast ? "Multicast" : "Unicast";
                const char *encap_name = "";
@@ -5430,7 +5442,7 @@ static int efx_ef10_filter_insert_def(struct efx_nic *efx,
                                           filter_flags, 0);
                        eth_broadcast_addr(baddr);
                        efx_filter_set_eth_local(&spec, vlan->vid, baddr);
-                       rc = efx_ef10_filter_insert(efx, &spec, true);
+                       rc = efx_ef10_filter_insert_locked(efx, &spec, true);
                        if (rc < 0) {
                                netif_warn(efx, drv, efx->net_dev,
                                           "Broadcast filter insert failed rc=%d\n",
index 570ec72266f37cac1626eb8358ffbf21bb536192..ce3a177081a854a683493f7f6f2c79ac63f60cc4 100644 (file)
@@ -1871,12 +1871,6 @@ static void efx_remove_filters(struct efx_nic *efx)
        up_write(&efx->filter_sem);
 }
 
-static void efx_restore_filters(struct efx_nic *efx)
-{
-       down_read(&efx->filter_sem);
-       efx->type->filter_table_restore(efx);
-       up_read(&efx->filter_sem);
-}
 
 /**************************************************************************
  *
@@ -2688,6 +2682,7 @@ void efx_reset_down(struct efx_nic *efx, enum reset_type method)
        efx_disable_interrupts(efx);
 
        mutex_lock(&efx->mac_lock);
+       down_write(&efx->filter_sem);
        mutex_lock(&efx->rss_lock);
        if (efx->port_initialized && method != RESET_TYPE_INVISIBLE &&
            method != RESET_TYPE_DATAPATH)
@@ -2745,9 +2740,8 @@ int efx_reset_up(struct efx_nic *efx, enum reset_type method, bool ok)
        if (efx->type->rx_restore_rss_contexts)
                efx->type->rx_restore_rss_contexts(efx);
        mutex_unlock(&efx->rss_lock);
-       down_read(&efx->filter_sem);
-       efx_restore_filters(efx);
-       up_read(&efx->filter_sem);
+       efx->type->filter_table_restore(efx);
+       up_write(&efx->filter_sem);
        if (efx->type->sriov_reset)
                efx->type->sriov_reset(efx);
 
@@ -2764,6 +2758,7 @@ fail:
        efx->port_initialized = false;
 
        mutex_unlock(&efx->rss_lock);
+       up_write(&efx->filter_sem);
        mutex_unlock(&efx->mac_lock);
 
        return rc;
@@ -3473,7 +3468,9 @@ static int efx_pci_probe_main(struct efx_nic *efx)
 
        efx_init_napi(efx);
 
+       down_write(&efx->filter_sem);
        rc = efx->type->init(efx);
+       up_write(&efx->filter_sem);
        if (rc) {
                netif_err(efx, probe, efx->net_dev,
                          "failed to initialise NIC\n");
@@ -3765,7 +3762,9 @@ static int efx_pm_resume(struct device *dev)
        rc = efx->type->reset(efx, RESET_TYPE_ALL);
        if (rc)
                return rc;
+       down_write(&efx->filter_sem);
        rc = efx->type->init(efx);
+       up_write(&efx->filter_sem);
        if (rc)
                return rc;
        rc = efx_pm_thaw(dev);
index 2e6e2a96b4f263023e04eaad77e56f160cbedc5c..f9a61f90cfbc6acb269d4e8320bb9a078ae04239 100644 (file)
@@ -37,7 +37,7 @@
  *             is done in the "stmmac files"
  */
 
-/* struct emac_variant - Descrive dwmac-sun8i hardware variant
+/* struct emac_variant - Describe dwmac-sun8i hardware variant
  * @default_syscon_value:      The default value of the EMAC register in syscon
  *                             This value is used for disabling properly EMAC
  *                             and used as a good starting value in case of the
index 6d141f3931eb650902469cebc18fa3613ad0dcb9..72da77b94ecd987e7e683d0ec890c842090e117e 100644 (file)
@@ -94,7 +94,6 @@ static int dwmac1000_validate_ucast_entries(int ucast_entries)
 /**
  * stmmac_axi_setup - parse DT parameters for programming the AXI register
  * @pdev: platform device
- * @priv: driver private struct.
  * Description:
  * if required, from device-tree the AXI internal register can be tuned
  * by using platform parameters.
index 8e9d0ee1572b97120546171f2487240e6e21bc43..31c3d77b4733f0aa9900138b5c49f398d0642db4 100644 (file)
@@ -1274,6 +1274,7 @@ int netvsc_poll(struct napi_struct *napi, int budget)
        struct hv_device *device = netvsc_channel_to_device(channel);
        struct net_device *ndev = hv_get_drvdata(device);
        int work_done = 0;
+       int ret;
 
        /* If starting a new interval */
        if (!nvchan->desc)
@@ -1285,16 +1286,18 @@ int netvsc_poll(struct napi_struct *napi, int budget)
                nvchan->desc = hv_pkt_iter_next(channel, nvchan->desc);
        }
 
-       /* If send of pending receive completions suceeded
-        *   and did not exhaust NAPI budget this time
-        *   and not doing busy poll
+       /* Send any pending receive completions */
+       ret = send_recv_completions(ndev, net_device, nvchan);
+
+       /* If it did not exhaust NAPI budget this time
+        *  and not doing busy poll
         * then re-enable host interrupts
-        *     and reschedule if ring is not empty.
+        *  and reschedule if ring is not empty
+        *   or sending receive completion failed.
         */
-       if (send_recv_completions(ndev, net_device, nvchan) == 0 &&
-           work_done < budget &&
+       if (work_done < budget &&
            napi_complete_done(napi, work_done) &&
-           hv_end_read(&channel->inbound) &&
+           (ret || hv_end_read(&channel->inbound)) &&
            napi_schedule_prep(napi)) {
                hv_begin_read(&channel->inbound);
                __napi_schedule(napi);
index 9b4e3c3787e5d33ff200a9711b1373b765b6052e..408ece27131c4611a8600028831f10aa8b47ed60 100644 (file)
@@ -1338,6 +1338,7 @@ out:
        /* setting up multiple channels failed */
        net_device->max_chn = 1;
        net_device->num_chn = 1;
+       return 0;
 
 err_dev_remv:
        rndis_filter_device_remove(dev, net_device);
index 64f1b1e77bc0f361dc59538072f59ca7a7c72690..23a52b9293f35eaec1d71063305a029ba466d819 100644 (file)
@@ -275,6 +275,8 @@ struct adf7242_local {
        struct spi_message stat_msg;
        struct spi_transfer stat_xfer;
        struct dentry *debugfs_root;
+       struct delayed_work work;
+       struct workqueue_struct *wqueue;
        unsigned long flags;
        int tx_stat;
        bool promiscuous;
@@ -575,10 +577,26 @@ static int adf7242_cmd_rx(struct adf7242_local *lp)
        /* Wait until the ACK is sent */
        adf7242_wait_status(lp, RC_STATUS_PHY_RDY, RC_STATUS_MASK, __LINE__);
        adf7242_clear_irqstat(lp);
+       mod_delayed_work(lp->wqueue, &lp->work, msecs_to_jiffies(400));
 
        return adf7242_cmd(lp, CMD_RC_RX);
 }
 
+static void adf7242_rx_cal_work(struct work_struct *work)
+{
+       struct adf7242_local *lp =
+       container_of(work, struct adf7242_local, work.work);
+
+       /* Reissuing RC_RX every 400ms - to adjust for offset
+        * drift in receiver (datasheet page 61, OCL section)
+        */
+
+       if (!test_bit(FLAG_XMIT, &lp->flags)) {
+               adf7242_cmd(lp, CMD_RC_PHY_RDY);
+               adf7242_cmd_rx(lp);
+       }
+}
+
 static int adf7242_set_txpower(struct ieee802154_hw *hw, int mbm)
 {
        struct adf7242_local *lp = hw->priv;
@@ -686,7 +704,7 @@ static int adf7242_start(struct ieee802154_hw *hw)
        enable_irq(lp->spi->irq);
        set_bit(FLAG_START, &lp->flags);
 
-       return adf7242_cmd(lp, CMD_RC_RX);
+       return adf7242_cmd_rx(lp);
 }
 
 static void adf7242_stop(struct ieee802154_hw *hw)
@@ -694,6 +712,7 @@ static void adf7242_stop(struct ieee802154_hw *hw)
        struct adf7242_local *lp = hw->priv;
 
        disable_irq(lp->spi->irq);
+       cancel_delayed_work_sync(&lp->work);
        adf7242_cmd(lp, CMD_RC_IDLE);
        clear_bit(FLAG_START, &lp->flags);
        adf7242_clear_irqstat(lp);
@@ -719,7 +738,10 @@ static int adf7242_channel(struct ieee802154_hw *hw, u8 page, u8 channel)
        adf7242_write_reg(lp, REG_CH_FREQ1, freq >> 8);
        adf7242_write_reg(lp, REG_CH_FREQ2, freq >> 16);
 
-       return adf7242_cmd(lp, CMD_RC_RX);
+       if (test_bit(FLAG_START, &lp->flags))
+               return adf7242_cmd_rx(lp);
+       else
+               return adf7242_cmd(lp, CMD_RC_PHY_RDY);
 }
 
 static int adf7242_set_hw_addr_filt(struct ieee802154_hw *hw,
@@ -814,6 +836,7 @@ static int adf7242_xmit(struct ieee802154_hw *hw, struct sk_buff *skb)
        /* ensure existing instances of the IRQ handler have completed */
        disable_irq(lp->spi->irq);
        set_bit(FLAG_XMIT, &lp->flags);
+       cancel_delayed_work_sync(&lp->work);
        reinit_completion(&lp->tx_complete);
        adf7242_cmd(lp, CMD_RC_PHY_RDY);
        adf7242_clear_irqstat(lp);
@@ -952,6 +975,7 @@ static irqreturn_t adf7242_isr(int irq, void *data)
        unsigned int xmit;
        u8 irq1;
 
+       mod_delayed_work(lp->wqueue, &lp->work, msecs_to_jiffies(400));
        adf7242_read_reg(lp, REG_IRQ1_SRC1, &irq1);
 
        if (!(irq1 & (IRQ_RX_PKT_RCVD | IRQ_CSMA_CA)))
@@ -1241,6 +1265,9 @@ static int adf7242_probe(struct spi_device *spi)
        spi_message_add_tail(&lp->stat_xfer, &lp->stat_msg);
 
        spi_set_drvdata(spi, lp);
+       INIT_DELAYED_WORK(&lp->work, adf7242_rx_cal_work);
+       lp->wqueue = alloc_ordered_workqueue(dev_name(&spi->dev),
+                                            WQ_MEM_RECLAIM);
 
        ret = adf7242_hw_init(lp);
        if (ret)
@@ -1284,6 +1311,9 @@ static int adf7242_remove(struct spi_device *spi)
        if (!IS_ERR_OR_NULL(lp->debugfs_root))
                debugfs_remove_recursive(lp->debugfs_root);
 
+       cancel_delayed_work_sync(&lp->work);
+       destroy_workqueue(lp->wqueue);
+
        ieee802154_unregister_hw(lp->hw);
        mutex_destroy(&lp->bmux);
        ieee802154_free_hw(lp->hw);
index 77abedf0b52447b4f1d0b5bdd99c259cb3555c1a..3d9e91579866826e476ceb2374b0d286e70c07fd 100644 (file)
@@ -940,7 +940,7 @@ at86rf230_xmit(struct ieee802154_hw *hw, struct sk_buff *skb)
 static int
 at86rf230_ed(struct ieee802154_hw *hw, u8 *level)
 {
-       BUG_ON(!level);
+       WARN_ON(!level);
        *level = 0xbe;
        return 0;
 }
@@ -1121,8 +1121,7 @@ at86rf230_set_hw_addr_filt(struct ieee802154_hw *hw,
        if (changed & IEEE802154_AFILT_SADDR_CHANGED) {
                u16 addr = le16_to_cpu(filt->short_addr);
 
-               dev_vdbg(&lp->spi->dev,
-                        "at86rf230_set_hw_addr_filt called for saddr\n");
+               dev_vdbg(&lp->spi->dev, "%s called for saddr\n", __func__);
                __at86rf230_write(lp, RG_SHORT_ADDR_0, addr);
                __at86rf230_write(lp, RG_SHORT_ADDR_1, addr >> 8);
        }
@@ -1130,8 +1129,7 @@ at86rf230_set_hw_addr_filt(struct ieee802154_hw *hw,
        if (changed & IEEE802154_AFILT_PANID_CHANGED) {
                u16 pan = le16_to_cpu(filt->pan_id);
 
-               dev_vdbg(&lp->spi->dev,
-                        "at86rf230_set_hw_addr_filt called for pan id\n");
+               dev_vdbg(&lp->spi->dev, "%s called for pan id\n", __func__);
                __at86rf230_write(lp, RG_PAN_ID_0, pan);
                __at86rf230_write(lp, RG_PAN_ID_1, pan >> 8);
        }
@@ -1140,15 +1138,13 @@ at86rf230_set_hw_addr_filt(struct ieee802154_hw *hw,
                u8 i, addr[8];
 
                memcpy(addr, &filt->ieee_addr, 8);
-               dev_vdbg(&lp->spi->dev,
-                        "at86rf230_set_hw_addr_filt called for IEEE addr\n");
+               dev_vdbg(&lp->spi->dev, "%s called for IEEE addr\n", __func__);
                for (i = 0; i < 8; i++)
                        __at86rf230_write(lp, RG_IEEE_ADDR_0 + i, addr[i]);
        }
 
        if (changed & IEEE802154_AFILT_PANC_CHANGED) {
-               dev_vdbg(&lp->spi->dev,
-                        "at86rf230_set_hw_addr_filt called for panc change\n");
+               dev_vdbg(&lp->spi->dev, "%s called for panc change\n", __func__);
                if (filt->pan_coord)
                        at86rf230_write_subreg(lp, SR_AACK_I_AM_COORD, 1);
                else
@@ -1252,7 +1248,6 @@ at86rf230_set_cca_mode(struct ieee802154_hw *hw,
        return at86rf230_write_subreg(lp, SR_CCA_MODE, val);
 }
 
-
 static int
 at86rf230_set_cca_ed_level(struct ieee802154_hw *hw, s32 mbm)
 {
index 0d673f7682ee065223b64462bc2c4df0a03826a0..176395e4b7bb0ca628bdd22b4f13a23e425bfae2 100644 (file)
@@ -49,7 +49,7 @@ struct fakelb_phy {
 
 static int fakelb_hw_ed(struct ieee802154_hw *hw, u8 *level)
 {
-       BUG_ON(!level);
+       WARN_ON(!level);
        *level = 0xbe;
 
        return 0;
index de0d7f28a181ca4acb1da2131d82a981627a8e96..e428277781ac4422bec2e8f47fd35476a85a74f7 100644 (file)
  */
 #include <linux/kernel.h>
 #include <linux/module.h>
-#include <linux/gpio.h>
+#include <linux/gpio/consumer.h>
 #include <linux/spi/spi.h>
 #include <linux/workqueue.h>
 #include <linux/interrupt.h>
+#include <linux/irq.h>
 #include <linux/skbuff.h>
 #include <linux/of_gpio.h>
 #include <linux/regmap.h>
index b8f57e9b937901fd142413c4002f39205546c35a..1cd439bdf6087af2913f589b499cd5c5abe5a3bb 100644 (file)
 #define MII_88E1318S_PHY_WOL_CTRL_CLEAR_WOL_STATUS             BIT(12)
 #define MII_88E1318S_PHY_WOL_CTRL_MAGIC_PACKET_MATCH_ENABLE    BIT(14)
 
-#define MII_88E1121_PHY_LED_CTRL       16
+#define MII_PHY_LED_CTRL               16
 #define MII_88E1121_PHY_LED_DEF                0x0030
+#define MII_88E1510_PHY_LED_DEF                0x1177
 
 #define MII_M1011_PHY_STATUS           0x11
 #define MII_M1011_PHY_STATUS_1000      0x8000
@@ -632,8 +633,40 @@ error:
        return err;
 }
 
+static void marvell_config_led(struct phy_device *phydev)
+{
+       u16 def_config;
+       int err;
+
+       switch (MARVELL_PHY_FAMILY_ID(phydev->phy_id)) {
+       /* Default PHY LED config: LED[0] .. Link, LED[1] .. Activity */
+       case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1121R):
+       case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1318S):
+               def_config = MII_88E1121_PHY_LED_DEF;
+               break;
+       /* Default PHY LED config:
+        * LED[0] .. 1000Mbps Link
+        * LED[1] .. 100Mbps Link
+        * LED[2] .. Blink, Activity
+        */
+       case MARVELL_PHY_FAMILY_ID(MARVELL_PHY_ID_88E1510):
+               def_config = MII_88E1510_PHY_LED_DEF;
+               break;
+       default:
+               return;
+       }
+
+       err = phy_write_paged(phydev, MII_MARVELL_LED_PAGE, MII_PHY_LED_CTRL,
+                             def_config);
+       if (err < 0)
+               pr_warn("Fail to config marvell phy LED.\n");
+}
+
 static int marvell_config_init(struct phy_device *phydev)
 {
+       /* Set defalut LED */
+       marvell_config_led(phydev);
+
        /* Set registers from marvell,reg-init DT property */
        return marvell_of_reg_init(phydev);
 }
@@ -813,21 +846,6 @@ static int m88e1111_config_init(struct phy_device *phydev)
        return genphy_soft_reset(phydev);
 }
 
-static int m88e1121_config_init(struct phy_device *phydev)
-{
-       int err;
-
-       /* Default PHY LED config: LED[0] .. Link, LED[1] .. Activity */
-       err = phy_write_paged(phydev, MII_MARVELL_LED_PAGE,
-                             MII_88E1121_PHY_LED_CTRL,
-                             MII_88E1121_PHY_LED_DEF);
-       if (err < 0)
-               return err;
-
-       /* Set marvell,reg-init configuration from device tree */
-       return marvell_config_init(phydev);
-}
-
 static int m88e1318_config_init(struct phy_device *phydev)
 {
        if (phy_interrupt_is_valid(phydev)) {
@@ -841,7 +859,7 @@ static int m88e1318_config_init(struct phy_device *phydev)
                        return err;
        }
 
-       return m88e1121_config_init(phydev);
+       return marvell_config_init(phydev);
 }
 
 static int m88e1510_config_init(struct phy_device *phydev)
@@ -2087,7 +2105,7 @@ static struct phy_driver marvell_drivers[] = {
                .features = PHY_GBIT_FEATURES,
                .flags = PHY_HAS_INTERRUPT,
                .probe = &m88e1121_probe,
-               .config_init = &m88e1121_config_init,
+               .config_init = &marvell_config_init,
                .config_aneg = &m88e1121_config_aneg,
                .read_status = &marvell_read_status,
                .ack_interrupt = &marvell_ack_interrupt,
index bd0f339f69fd064737f8f3c80e6645e73c56a2b9..b9f5f40a7ac1e6640a653e8207cdd8885100e09f 100644 (file)
@@ -1724,11 +1724,8 @@ EXPORT_SYMBOL(genphy_loopback);
 
 static int __set_phy_supported(struct phy_device *phydev, u32 max_speed)
 {
-       /* The default values for phydev->supported are provided by the PHY
-        * driver "features" member, we want to reset to sane defaults first
-        * before supporting higher speeds.
-        */
-       phydev->supported &= PHY_DEFAULT_FEATURES;
+       phydev->supported &= ~(PHY_1000BT_FEATURES | PHY_100BT_FEATURES |
+                              PHY_10BT_FEATURES);
 
        switch (max_speed) {
        default:
index d437f4f5ed5291d21236a71ef3e36089344f9201..740655261e5b7347116d2a5b53445c8d023cb49c 100644 (file)
@@ -349,7 +349,6 @@ static int sfp_register_bus(struct sfp_bus *bus)
        }
        if (bus->started)
                bus->socket_ops->start(bus->sfp);
-       bus->netdev->sfp_bus = bus;
        bus->registered = true;
        return 0;
 }
@@ -364,7 +363,6 @@ static void sfp_unregister_bus(struct sfp_bus *bus)
                if (bus->phydev && ops && ops->disconnect_phy)
                        ops->disconnect_phy(bus->upstream);
        }
-       bus->netdev->sfp_bus = NULL;
        bus->registered = false;
 }
 
@@ -436,6 +434,14 @@ void sfp_upstream_stop(struct sfp_bus *bus)
 }
 EXPORT_SYMBOL_GPL(sfp_upstream_stop);
 
+static void sfp_upstream_clear(struct sfp_bus *bus)
+{
+       bus->upstream_ops = NULL;
+       bus->upstream = NULL;
+       bus->netdev->sfp_bus = NULL;
+       bus->netdev = NULL;
+}
+
 /**
  * sfp_register_upstream() - Register the neighbouring device
  * @fwnode: firmware node for the SFP bus
@@ -461,9 +467,13 @@ struct sfp_bus *sfp_register_upstream(struct fwnode_handle *fwnode,
                bus->upstream_ops = ops;
                bus->upstream = upstream;
                bus->netdev = ndev;
+               ndev->sfp_bus = bus;
 
-               if (bus->sfp)
+               if (bus->sfp) {
                        ret = sfp_register_bus(bus);
+                       if (ret)
+                               sfp_upstream_clear(bus);
+               }
                rtnl_unlock();
        }
 
@@ -488,8 +498,7 @@ void sfp_unregister_upstream(struct sfp_bus *bus)
        rtnl_lock();
        if (bus->sfp)
                sfp_unregister_bus(bus);
-       bus->upstream = NULL;
-       bus->netdev = NULL;
+       sfp_upstream_clear(bus);
        rtnl_unlock();
 
        sfp_bus_put(bus);
@@ -561,6 +570,13 @@ void sfp_module_remove(struct sfp_bus *bus)
 }
 EXPORT_SYMBOL_GPL(sfp_module_remove);
 
+static void sfp_socket_clear(struct sfp_bus *bus)
+{
+       bus->sfp_dev = NULL;
+       bus->sfp = NULL;
+       bus->socket_ops = NULL;
+}
+
 struct sfp_bus *sfp_register_socket(struct device *dev, struct sfp *sfp,
                                    const struct sfp_socket_ops *ops)
 {
@@ -573,8 +589,11 @@ struct sfp_bus *sfp_register_socket(struct device *dev, struct sfp *sfp,
                bus->sfp = sfp;
                bus->socket_ops = ops;
 
-               if (bus->netdev)
+               if (bus->netdev) {
                        ret = sfp_register_bus(bus);
+                       if (ret)
+                               sfp_socket_clear(bus);
+               }
                rtnl_unlock();
        }
 
@@ -592,9 +611,7 @@ void sfp_unregister_socket(struct sfp_bus *bus)
        rtnl_lock();
        if (bus->netdev)
                sfp_unregister_bus(bus);
-       bus->sfp_dev = NULL;
-       bus->sfp = NULL;
-       bus->socket_ops = NULL;
+       sfp_socket_clear(bus);
        rtnl_unlock();
 
        sfp_bus_put(bus);
index a192a017cc68878360505b93df151de3d0b9b730..f5727baac84a5d10fd70837a75fcfa8194992f9a 100644 (file)
@@ -1688,7 +1688,7 @@ static struct sk_buff *tun_build_skb(struct tun_struct *tun,
                case XDP_TX:
                        get_page(alloc_frag->page);
                        alloc_frag->offset += buflen;
-                       if (tun_xdp_tx(tun->dev, &xdp))
+                       if (tun_xdp_tx(tun->dev, &xdp) < 0)
                                goto err_redirect;
                        rcu_read_unlock();
                        local_bh_enable();
index 3d4f7959dabb9c39e17754df4f72013c89743d5a..b1b3d8f7e67dd052eae618e33698c633751df60a 100644 (file)
@@ -642,10 +642,12 @@ static void ax88772_restore_phy(struct usbnet *dev)
                                     priv->presvd_phy_advertise);
 
                /* Restore BMCR */
+               if (priv->presvd_phy_bmcr & BMCR_ANENABLE)
+                       priv->presvd_phy_bmcr |= BMCR_ANRESTART;
+
                asix_mdio_write_nopm(dev->net, dev->mii.phy_id, MII_BMCR,
                                     priv->presvd_phy_bmcr);
 
-               mii_nway_restart(&dev->mii);
                priv->presvd_phy_advertise = 0;
                priv->presvd_phy_bmcr = 0;
        }
index 2e4130746c40caf6a2a5dbf51fd0ece84fb6da85..ed10d49eb5e0b66068fe366950cee3de2de3257d 100644 (file)
@@ -3344,6 +3344,7 @@ static void lan78xx_tx_bh(struct lan78xx_net *dev)
        pkt_cnt = 0;
        count = 0;
        length = 0;
+       spin_lock_irqsave(&tqp->lock, flags);
        for (skb = tqp->next; pkt_cnt < tqp->qlen; skb = skb->next) {
                if (skb_is_gso(skb)) {
                        if (pkt_cnt) {
@@ -3352,7 +3353,8 @@ static void lan78xx_tx_bh(struct lan78xx_net *dev)
                        }
                        count = 1;
                        length = skb->len - TX_OVERHEAD;
-                       skb2 = skb_dequeue(tqp);
+                       __skb_unlink(skb, tqp);
+                       spin_unlock_irqrestore(&tqp->lock, flags);
                        goto gso_skb;
                }
 
@@ -3361,6 +3363,7 @@ static void lan78xx_tx_bh(struct lan78xx_net *dev)
                skb_totallen = skb->len + roundup(skb_totallen, sizeof(u32));
                pkt_cnt++;
        }
+       spin_unlock_irqrestore(&tqp->lock, flags);
 
        /* copy to a single skb */
        skb = alloc_skb(skb_totallen, GFP_ATOMIC);
index 8fac8e132c5bd250f1a528a020adeb6eb2f75cd1..38502809420b9c76ac5fdc89c59faf71dff41e9b 100644 (file)
@@ -1253,6 +1253,7 @@ static const struct usb_device_id products[] = {
        {QMI_QUIRK_SET_DTR(0x1e0e, 0x9001, 5)}, /* SIMCom 7100E, 7230E, 7600E ++ */
        {QMI_QUIRK_SET_DTR(0x2c7c, 0x0125, 4)}, /* Quectel EC25, EC20 R2.0  Mini PCIe */
        {QMI_QUIRK_SET_DTR(0x2c7c, 0x0121, 4)}, /* Quectel EC21 Mini PCIe */
+       {QMI_QUIRK_SET_DTR(0x2c7c, 0x0191, 4)}, /* Quectel EG91 */
        {QMI_FIXED_INTF(0x2c7c, 0x0296, 4)},    /* Quectel BG96 */
        {QMI_QUIRK_SET_DTR(0x2c7c, 0x0306, 4)}, /* Quectel EP06 Mini PCIe */
 
index 5f565bd574da3bc7ce741e3b280a9ff5dece4352..48ba80a8ca5ce8e566931979edcff4bcfe47bc2e 100644 (file)
@@ -681,7 +681,7 @@ static void rtl8150_set_multicast(struct net_device *netdev)
                   (netdev->flags & IFF_ALLMULTI)) {
                rx_creg &= 0xfffe;
                rx_creg |= 0x0002;
-               dev_info(&netdev->dev, "%s: allmulti set\n", netdev->name);
+               dev_dbg(&netdev->dev, "%s: allmulti set\n", netdev->name);
        } else {
                /* ~RX_MULTICAST, ~RX_PROMISCUOUS */
                rx_creg &= 0x00fc;
index 7a6a1fe793090b8e28f5ef075f5ebc2ad385b5eb..05553d2524469f97e4a02bb48f43f6820ad2b3e5 100644 (file)
@@ -82,6 +82,9 @@ static bool turbo_mode = true;
 module_param(turbo_mode, bool, 0644);
 MODULE_PARM_DESC(turbo_mode, "Enable multiple frames per Rx transaction");
 
+static int smsc75xx_link_ok_nopm(struct usbnet *dev);
+static int smsc75xx_phy_gig_workaround(struct usbnet *dev);
+
 static int __must_check __smsc75xx_read_reg(struct usbnet *dev, u32 index,
                                            u32 *data, int in_pm)
 {
@@ -852,6 +855,9 @@ static int smsc75xx_phy_initialize(struct usbnet *dev)
                return -EIO;
        }
 
+       /* phy workaround for gig link */
+       smsc75xx_phy_gig_workaround(dev);
+
        smsc75xx_mdio_write(dev->net, dev->mii.phy_id, MII_ADVERTISE,
                ADVERTISE_ALL | ADVERTISE_CSMA | ADVERTISE_PAUSE_CAP |
                ADVERTISE_PAUSE_ASYM);
@@ -987,6 +993,62 @@ static int smsc75xx_wait_ready(struct usbnet *dev, int in_pm)
        return -EIO;
 }
 
+static int smsc75xx_phy_gig_workaround(struct usbnet *dev)
+{
+       struct mii_if_info *mii = &dev->mii;
+       int ret = 0, timeout = 0;
+       u32 buf, link_up = 0;
+
+       /* Set the phy in Gig loopback */
+       smsc75xx_mdio_write(dev->net, mii->phy_id, MII_BMCR, 0x4040);
+
+       /* Wait for the link up */
+       do {
+               link_up = smsc75xx_link_ok_nopm(dev);
+               usleep_range(10000, 20000);
+               timeout++;
+       } while ((!link_up) && (timeout < 1000));
+
+       if (timeout >= 1000) {
+               netdev_warn(dev->net, "Timeout waiting for PHY link up\n");
+               return -EIO;
+       }
+
+       /* phy reset */
+       ret = smsc75xx_read_reg(dev, PMT_CTL, &buf);
+       if (ret < 0) {
+               netdev_warn(dev->net, "Failed to read PMT_CTL: %d\n", ret);
+               return ret;
+       }
+
+       buf |= PMT_CTL_PHY_RST;
+
+       ret = smsc75xx_write_reg(dev, PMT_CTL, buf);
+       if (ret < 0) {
+               netdev_warn(dev->net, "Failed to write PMT_CTL: %d\n", ret);
+               return ret;
+       }
+
+       timeout = 0;
+       do {
+               usleep_range(10000, 20000);
+               ret = smsc75xx_read_reg(dev, PMT_CTL, &buf);
+               if (ret < 0) {
+                       netdev_warn(dev->net, "Failed to read PMT_CTL: %d\n",
+                                   ret);
+                       return ret;
+               }
+               timeout++;
+       } while ((buf & PMT_CTL_PHY_RST) && (timeout < 100));
+
+       if (timeout >= 100) {
+               netdev_warn(dev->net, "timeout waiting for PHY Reset\n");
+               return -EIO;
+       }
+
+       return 0;
+}
+
 static int smsc75xx_reset(struct usbnet *dev)
 {
        struct smsc75xx_priv *pdata = (struct smsc75xx_priv *)(dev->data[0]);
index e9c2fb318c03362d84031241a4191db9f4602c1a..836e0a47b94a0a192b210620d6652c41145cbab1 100644 (file)
@@ -6058,8 +6058,19 @@ static void ath10k_sta_rc_update_wk(struct work_struct *wk)
                           ath10k_mac_max_vht_nss(vht_mcs_mask)));
 
        if (changed & IEEE80211_RC_BW_CHANGED) {
-               ath10k_dbg(ar, ATH10K_DBG_MAC, "mac update sta %pM peer bw %d\n",
-                          sta->addr, bw);
+               enum wmi_phy_mode mode;
+
+               mode = chan_to_phymode(&def);
+               ath10k_dbg(ar, ATH10K_DBG_MAC, "mac update sta %pM peer bw %d phymode %d\n",
+                               sta->addr, bw, mode);
+
+               err = ath10k_wmi_peer_set_param(ar, arvif->vdev_id, sta->addr,
+                               WMI_PEER_PHYMODE, mode);
+               if (err) {
+                       ath10k_warn(ar, "failed to update STA %pM peer phymode %d: %d\n",
+                                       sta->addr, mode, err);
+                       goto exit;
+               }
 
                err = ath10k_wmi_peer_set_param(ar, arvif->vdev_id, sta->addr,
                                                WMI_PEER_CHAN_WIDTH, bw);
@@ -6100,6 +6111,7 @@ static void ath10k_sta_rc_update_wk(struct work_struct *wk)
                                    sta->addr);
        }
 
+exit:
        mutex_unlock(&ar->conf_mutex);
 }
 
index b48db54e986516271daab61f30277815a7a9bf6d..d68afb65402a069528b0dc3a01eab142c833898c 100644 (file)
@@ -6144,6 +6144,7 @@ enum wmi_peer_param {
        WMI_PEER_NSS        = 0x5,
        WMI_PEER_USE_4ADDR  = 0x6,
        WMI_PEER_DEBUG      = 0xa,
+       WMI_PEER_PHYMODE    = 0xd,
        WMI_PEER_DUMMY_VAR  = 0xff, /* dummy parameter for STA PS workaround */
 };
 
index 1279064a3b716c2ef6cf82d82c27ea664f1496b4..51a038022c8b80404b9bd841c6fefd3b866ffe66 100644 (file)
@@ -1,4 +1,4 @@
-/*
+/*
  * Copyright (c) 2018, The Linux Foundation. All rights reserved.
  *
  * Permission to use, copy, modify, and/or distribute this software for any
index c99a191e8d693a3e6ef006826fcde5affb74a02d..a907d7b065fa8e0b7ab6a35dc2c265a6385d9c75 100644 (file)
@@ -4296,6 +4296,13 @@ void brcmf_sdio_remove(struct brcmf_sdio *bus)
        brcmf_dbg(TRACE, "Enter\n");
 
        if (bus) {
+               /* Stop watchdog task */
+               if (bus->watchdog_tsk) {
+                       send_sig(SIGTERM, bus->watchdog_tsk, 1);
+                       kthread_stop(bus->watchdog_tsk);
+                       bus->watchdog_tsk = NULL;
+               }
+
                /* De-register interrupt handler */
                brcmf_sdiod_intr_unregister(bus->sdiodev);
 
index 6e3cf9817730b53f751f31401425ffd1c1dc82a4..88f4c89f89ba85f5ff64085f8f29abf20722d5ea 100644 (file)
@@ -644,11 +644,6 @@ static void mwifiex_usb_disconnect(struct usb_interface *intf)
                                         MWIFIEX_FUNC_SHUTDOWN);
        }
 
-       if (adapter->workqueue)
-               flush_workqueue(adapter->workqueue);
-
-       mwifiex_usb_free(card);
-
        mwifiex_dbg(adapter, FATAL,
                    "%s: removing card\n", __func__);
        mwifiex_remove_card(adapter);
@@ -1356,6 +1351,8 @@ static void mwifiex_unregister_dev(struct mwifiex_adapter *adapter)
 {
        struct usb_card_rec *card = (struct usb_card_rec *)adapter->card;
 
+       mwifiex_usb_free(card);
+
        mwifiex_usb_cleanup_tx_aggr(adapter);
 
        card->adapter = NULL;
index 9d2f9a776ef18e405c80e73c40c9eb3bc137b50d..b804abd464ae06365adbe108a5706412efe53f4b 100644 (file)
@@ -986,13 +986,15 @@ static void mt7601u_agc_tune(struct mt7601u_dev *dev)
         */
        spin_lock_bh(&dev->con_mon_lock);
        avg_rssi = ewma_rssi_read(&dev->avg_rssi);
-       WARN_ON_ONCE(avg_rssi == 0);
+       spin_unlock_bh(&dev->con_mon_lock);
+       if (avg_rssi == 0)
+               return;
+
        avg_rssi = -avg_rssi;
        if (avg_rssi <= -70)
                val -= 0x20;
        else if (avg_rssi <= -60)
                val -= 0x10;
-       spin_unlock_bh(&dev->con_mon_lock);
 
        if (val != mt7601u_bbp_rr(dev, 66))
                mt7601u_bbp_wr(dev, 66, val);
index 220e2b71020859163cc4affc71f505648561b151..ae0ca800684950e65ecc01916f4782af54a8e0eb 100644 (file)
@@ -654,8 +654,7 @@ qtnf_disconnect(struct wiphy *wiphy, struct net_device *dev,
        vif = qtnf_mac_get_base_vif(mac);
        if (!vif) {
                pr_err("MAC%u: primary VIF is not configured\n", mac->macid);
-               ret = -EFAULT;
-               goto out;
+               return -EFAULT;
        }
 
        if (vif->wdev.iftype != NL80211_IFTYPE_STATION) {
index 39c817eddd78e9cf736fbbd440c6617867afad20..54c9f6ab0c8cadb483d10413783b45b401c6f6f4 100644 (file)
@@ -484,18 +484,21 @@ static void _rtl_init_deferred_work(struct ieee80211_hw *hw)
 
 }
 
-void rtl_deinit_deferred_work(struct ieee80211_hw *hw)
+void rtl_deinit_deferred_work(struct ieee80211_hw *hw, bool ips_wq)
 {
        struct rtl_priv *rtlpriv = rtl_priv(hw);
 
        del_timer_sync(&rtlpriv->works.watchdog_timer);
 
-       cancel_delayed_work(&rtlpriv->works.watchdog_wq);
-       cancel_delayed_work(&rtlpriv->works.ips_nic_off_wq);
-       cancel_delayed_work(&rtlpriv->works.ps_work);
-       cancel_delayed_work(&rtlpriv->works.ps_rfon_wq);
-       cancel_delayed_work(&rtlpriv->works.fwevt_wq);
-       cancel_delayed_work(&rtlpriv->works.c2hcmd_wq);
+       cancel_delayed_work_sync(&rtlpriv->works.watchdog_wq);
+       if (ips_wq)
+               cancel_delayed_work(&rtlpriv->works.ips_nic_off_wq);
+       else
+               cancel_delayed_work_sync(&rtlpriv->works.ips_nic_off_wq);
+       cancel_delayed_work_sync(&rtlpriv->works.ps_work);
+       cancel_delayed_work_sync(&rtlpriv->works.ps_rfon_wq);
+       cancel_delayed_work_sync(&rtlpriv->works.fwevt_wq);
+       cancel_delayed_work_sync(&rtlpriv->works.c2hcmd_wq);
 }
 EXPORT_SYMBOL_GPL(rtl_deinit_deferred_work);
 
index 912f205779c39e68387269825fdae5c4fcaa206f..a7ae40eaa3cd538f96622e4e9a53da3c2b13ccec 100644 (file)
@@ -121,7 +121,7 @@ void rtl_init_rfkill(struct ieee80211_hw *hw);
 void rtl_deinit_rfkill(struct ieee80211_hw *hw);
 
 void rtl_watch_dog_timer_callback(struct timer_list *t);
-void rtl_deinit_deferred_work(struct ieee80211_hw *hw);
+void rtl_deinit_deferred_work(struct ieee80211_hw *hw, bool ips_wq);
 
 bool rtl_action_proc(struct ieee80211_hw *hw, struct sk_buff *skb, u8 is_tx);
 int rtlwifi_rate_mapping(struct ieee80211_hw *hw, bool isht,
index cfea57efa7f43c6bc1c6e99a45978053a1eed2ba..4bf7967590ca7be3b9b452d0b0ec65d98b829599 100644 (file)
@@ -130,7 +130,6 @@ found_alt:
                       firmware->size);
                rtlpriv->rtlhal.wowlan_fwsize = firmware->size;
        }
-       rtlpriv->rtlhal.fwsize = firmware->size;
        release_firmware(firmware);
 }
 
@@ -196,7 +195,7 @@ static void rtl_op_stop(struct ieee80211_hw *hw)
                /* reset sec info */
                rtl_cam_reset_sec_info(hw);
 
-               rtl_deinit_deferred_work(hw);
+               rtl_deinit_deferred_work(hw, false);
        }
        rtlpriv->intf_ops->adapter_stop(hw);
 
index ae13bcfb3bf09cc142a81c9ab78c31ad80fd035e..5d1fda16fc8c4c966ff8e24ca8d49bd3b6ca38c6 100644 (file)
@@ -2377,7 +2377,7 @@ void rtl_pci_disconnect(struct pci_dev *pdev)
                ieee80211_unregister_hw(hw);
                rtlmac->mac80211_registered = 0;
        } else {
-               rtl_deinit_deferred_work(hw);
+               rtl_deinit_deferred_work(hw, false);
                rtlpriv->intf_ops->adapter_stop(hw);
        }
        rtlpriv->cfg->ops->disable_interrupt(hw);
index 71af24e2e05197a344dd549c94d6f8ecba7ce00f..479a4cfc245d349e105457845719ca79b7ebb10d 100644 (file)
@@ -71,7 +71,7 @@ bool rtl_ps_disable_nic(struct ieee80211_hw *hw)
        struct rtl_priv *rtlpriv = rtl_priv(hw);
 
        /*<1> Stop all timer */
-       rtl_deinit_deferred_work(hw);
+       rtl_deinit_deferred_work(hw, true);
 
        /*<2> Disable Interrupt */
        rtlpriv->cfg->ops->disable_interrupt(hw);
@@ -292,7 +292,7 @@ void rtl_ips_nic_on(struct ieee80211_hw *hw)
        struct rtl_ps_ctl *ppsc = rtl_psc(rtl_priv(hw));
        enum rf_pwrstate rtstate;
 
-       cancel_delayed_work(&rtlpriv->works.ips_nic_off_wq);
+       cancel_delayed_work_sync(&rtlpriv->works.ips_nic_off_wq);
 
        mutex_lock(&rtlpriv->locks.ips_mutex);
        if (ppsc->inactiveps) {
index f9faffc498bcbd2d94cad365814955f9b1347759..2ac5004d7a401ab5d1255126c5c0a00a5e233705 100644 (file)
@@ -1132,7 +1132,7 @@ void rtl_usb_disconnect(struct usb_interface *intf)
                ieee80211_unregister_hw(hw);
                rtlmac->mac80211_registered = 0;
        } else {
-               rtl_deinit_deferred_work(hw);
+               rtl_deinit_deferred_work(hw, false);
                rtlpriv->intf_ops->adapter_stop(hw);
        }
        /*deinit rfkill */
index 547dbdac9d541ebaf57883eaa07e50786f2e46e4..01b0e2bb33190c78fb3818e34d5aebf4f60b2832 100644 (file)
@@ -89,6 +89,7 @@ int ptp_set_pinfunc(struct ptp_clock *ptp, unsigned int pin,
        case PTP_PF_PHYSYNC:
                if (chan != 0)
                        return -EINVAL;
+               break;
        default:
                return -EINVAL;
        }
index 79795c5fa7c37d3281fdc006ac701a4cb17c4cf6..d50c2f0a655ae3f95271d5f8de40f8eabc917c65 100644 (file)
@@ -2,6 +2,7 @@
 #ifndef _BPF_CGROUP_H
 #define _BPF_CGROUP_H
 
+#include <linux/errno.h>
 #include <linux/jump_label.h>
 #include <uapi/linux/bpf.h>
 
index 300baad62c889100722b1778a7c4819a16fbe2d1..c73dd7396886751938a0e2e1355d2aa28797ad87 100644 (file)
@@ -765,8 +765,8 @@ static inline bool bpf_dump_raw_ok(void)
 struct bpf_prog *bpf_patch_insn_single(struct bpf_prog *prog, u32 off,
                                       const struct bpf_insn *patch, u32 len);
 
-static inline int __xdp_generic_ok_fwd_dev(struct sk_buff *skb,
-                                          struct net_device *fwd)
+static inline int xdp_ok_fwd_dev(const struct net_device *fwd,
+                                unsigned int pktlen)
 {
        unsigned int len;
 
@@ -774,7 +774,7 @@ static inline int __xdp_generic_ok_fwd_dev(struct sk_buff *skb,
                return -ENETDOWN;
 
        len = fwd->mtu + fwd->hard_header_len + VLAN_HLEN;
-       if (skb->len > len)
+       if (pktlen > len)
                return -EMSGSIZE;
 
        return 0;
index 3efa3b861d44cae46670532c9db208d8630099a9..941b11811f85915bd70a730bbc338288d995493b 100644 (file)
@@ -16,6 +16,7 @@
 #define __FSL_GUTS_H__
 
 #include <linux/types.h>
+#include <linux/io.h>
 
 /**
  * Global Utility Registers.
index 7843b98e1c6ea7802dcea3f8b5a944d2355398d5..c20c7e197d0731e58b0f68b87531299080e8421a 100644 (file)
@@ -105,13 +105,13 @@ static inline bool br_vlan_enabled(const struct net_device *dev)
 
 static inline int br_vlan_get_pvid(const struct net_device *dev, u16 *p_pvid)
 {
-       return -1;
+       return -EINVAL;
 }
 
 static inline int br_vlan_get_info(const struct net_device *dev, u16 vid,
                                   struct bridge_vlan_info *p_vinfo)
 {
-       return -1;
+       return -EINVAL;
 }
 #endif
 
index f8231854b5d60316310fc5d8e57eea8625fe3078..119f53941c124c22452bf615f9ccca5a9130bb87 100644 (file)
@@ -109,6 +109,8 @@ struct ip_mc_list {
 extern int ip_check_mc_rcu(struct in_device *dev, __be32 mc_addr, __be32 src_addr, u8 proto);
 extern int igmp_rcv(struct sk_buff *);
 extern int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr);
+extern int ip_mc_join_group_ssm(struct sock *sk, struct ip_mreqn *imr,
+                               unsigned int mode);
 extern int ip_mc_leave_group(struct sock *sk, struct ip_mreqn *imr);
 extern void ip_mc_drop_socket(struct sock *sk);
 extern int ip_mc_source(int add, int omode, struct sock *sk,
index 4f5f8c21e2830bd3c7de20bd509b360f27e66816..1eb6f244588dae1efa08a2c9dcb8e46460592bcc 100644 (file)
@@ -27,6 +27,8 @@
  */
 #define MARVELL_PHY_ID_88E6390         0x01410f90
 
+#define MARVELL_PHY_FAMILY_ID(id)      ((id) >> 4)
+
 /* struct phy_device dev_flags definitions */
 #define MARVELL_PHY_M1145_FLAGS_RESISTANCE     0x00000001
 #define MARVELL_PHY_M1118_DNS323_LEDS          0x00000002
index 164cdedf6012c1431c8e3324a5029c2cb1abe141..610a201126ee031166798baaf8ecae74fe478c4d 100644 (file)
@@ -630,6 +630,7 @@ typedef unsigned char *sk_buff_data_t;
  *     @hash: the packet hash
  *     @queue_mapping: Queue mapping for multiqueue devices
  *     @xmit_more: More SKBs are pending for this queue
+ *     @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
  *     @ndisc_nodetype: router type (from link layer)
  *     @ooo_okay: allow the mapping of a socket to a queue to be changed
  *     @l4_hash: indicate hash is a canonical 4-tuple hash over transport
@@ -735,7 +736,7 @@ struct sk_buff {
                                peeked:1,
                                head_frag:1,
                                xmit_more:1,
-                               __unused:1; /* one bit hole */
+                               pfmemalloc:1;
 
        /* fields enclosed in headers_start/headers_end are copied
         * using a single memcpy() in __copy_skb_header()
@@ -754,31 +755,30 @@ struct sk_buff {
 
        __u8                    __pkt_type_offset[0];
        __u8                    pkt_type:3;
-       __u8                    pfmemalloc:1;
        __u8                    ignore_df:1;
-
        __u8                    nf_trace:1;
        __u8                    ip_summed:2;
        __u8                    ooo_okay:1;
+
        __u8                    l4_hash:1;
        __u8                    sw_hash:1;
        __u8                    wifi_acked_valid:1;
        __u8                    wifi_acked:1;
-
        __u8                    no_fcs:1;
        /* Indicates the inner headers are valid in the skbuff. */
        __u8                    encapsulation:1;
        __u8                    encap_hdr_csum:1;
        __u8                    csum_valid:1;
+
        __u8                    csum_complete_sw:1;
        __u8                    csum_level:2;
        __u8                    csum_not_inet:1;
-
        __u8                    dst_pending_confirm:1;
 #ifdef CONFIG_IPV6_NDISC_NODETYPE
        __u8                    ndisc_nodetype:2;
 #endif
        __u8                    ipvs_property:1;
+
        __u8                    inner_protocol_type:1;
        __u8                    remcsum_offload:1;
 #ifdef CONFIG_NET_SWITCHDEV
index 59656fc580df7e0301e0c9282af9358a255b863f..7b9c82de11cc9388b070992af610e5fd14b66333 100644 (file)
@@ -66,6 +66,12 @@ static inline bool rt6_need_strict(const struct in6_addr *daddr)
                (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
 }
 
+static inline bool rt6_qualify_for_ecmp(const struct fib6_info *f6i)
+{
+       return (f6i->fib6_flags & (RTF_GATEWAY|RTF_ADDRCONF|RTF_DYNAMIC)) ==
+              RTF_GATEWAY;
+}
+
 void ip6_route_input(struct sk_buff *skb);
 struct dst_entry *ip6_route_input_lookup(struct net *net,
                                         struct net_device *dev,
index 16475c269749a72f3c487e102e50cabff797317e..8f73be4945037c6d0997ec8ab7c3e9da3980a6e4 100644 (file)
@@ -355,14 +355,7 @@ struct ipv6_txoptions *ipv6_dup_options(struct sock *sk,
 struct ipv6_txoptions *ipv6_renew_options(struct sock *sk,
                                          struct ipv6_txoptions *opt,
                                          int newtype,
-                                         struct ipv6_opt_hdr __user *newopt,
-                                         int newoptlen);
-struct ipv6_txoptions *
-ipv6_renew_options_kern(struct sock *sk,
-                       struct ipv6_txoptions *opt,
-                       int newtype,
-                       struct ipv6_opt_hdr *newopt,
-                       int newoptlen);
+                                         struct ipv6_opt_hdr *newopt);
 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
                                          struct ipv6_txoptions *opt);
 
@@ -830,7 +823,7 @@ static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
         * to minimize possbility that any useful information to an
         * attacker is leaked. Only lower 20 bits are relevant.
         */
-       rol32(hash, 16);
+       hash = rol32(hash, 16);
 
        flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
 
@@ -1107,6 +1100,8 @@ void ipv6_sysctl_unregister(void);
 
 int ipv6_sock_mc_join(struct sock *sk, int ifindex,
                      const struct in6_addr *addr);
+int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
+                         const struct in6_addr *addr, unsigned int mode);
 int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
                      const struct in6_addr *addr);
 #endif /* _NET_IPV6_H */
index e0c0c2558ec48adfb27629c2180f9b04efb67bcf..a05134507e7bc806d9afd9ff7c86b95e5df084eb 100644 (file)
@@ -65,4 +65,10 @@ extern const struct nft_expr_ops nft_payload_fast_ops;
 extern struct static_key_false nft_counters_enabled;
 extern struct static_key_false nft_trace_enabled;
 
+extern struct nft_set_type nft_set_rhash_type;
+extern struct nft_set_type nft_set_hash_type;
+extern struct nft_set_type nft_set_hash_fast_type;
+extern struct nft_set_type nft_set_rbtree_type;
+extern struct nft_set_type nft_set_bitmap_type;
+
 #endif /* _NET_NF_TABLES_CORE_H */
index 9754a50ecde9c44162cc60e387d48cb034c6e6d4..4cc64c8446eb94f1c122cf15d4bf74c7e3f2275d 100644 (file)
@@ -64,7 +64,7 @@ nf_tproxy_handle_time_wait4(struct net *net, struct sk_buff *skb,
  * belonging to established connections going through that one.
  */
 struct sock *
-nf_tproxy_get_sock_v4(struct net *net, struct sk_buff *skb, void *hp,
+nf_tproxy_get_sock_v4(struct net *net, struct sk_buff *skb,
                      const u8 protocol,
                      const __be32 saddr, const __be32 daddr,
                      const __be16 sport, const __be16 dport,
@@ -103,7 +103,7 @@ nf_tproxy_handle_time_wait6(struct sk_buff *skb, int tproto, int thoff,
                            struct sock *sk);
 
 struct sock *
-nf_tproxy_get_sock_v6(struct net *net, struct sk_buff *skb, int thoff, void *hp,
+nf_tproxy_get_sock_v6(struct net *net, struct sk_buff *skb, int thoff,
                      const u8 protocol,
                      const struct in6_addr *saddr, const struct in6_addr *daddr,
                      const __be16 sport, const __be16 dport,
index 9470fd7e4350ea9546b43a504ebe12f6362dda18..32d2454c04793021c0dc87bca7f1802b49c5249b 100644 (file)
@@ -7,7 +7,6 @@
 #include <linux/tc_act/tc_csum.h>
 
 struct tcf_csum_params {
-       int action;
        u32 update_flags;
        struct rcu_head rcu;
 };
index efef0b4b1b2bddc76095bcd4d02ebaaa3b2beb56..46b8c7f1c8d5273791df55eeb6345807d8812e96 100644 (file)
@@ -18,7 +18,6 @@
 struct tcf_tunnel_key_params {
        struct rcu_head         rcu;
        int                     tcft_action;
-       int                     action;
        struct metadata_dst     *tcft_enc_metadata;
 };
 
index 800582b5dd54f46dc2a853f60994e26562ae1b28..3482d13d655b88910e3d41d30584d94bfea7a535 100644 (file)
@@ -828,6 +828,10 @@ struct tcp_skb_cb {
 
 #define TCP_SKB_CB(__skb)      ((struct tcp_skb_cb *)&((__skb)->cb[0]))
 
+static inline void bpf_compute_data_end_sk_skb(struct sk_buff *skb)
+{
+       TCP_SKB_CB(skb)->bpf.data_end = skb->data + skb_headlen(skb);
+}
 
 #if IS_ENABLED(CONFIG_IPV6)
 /* This is the variant of inet6_iif() that must be used by TCP,
@@ -908,8 +912,6 @@ enum tcp_ca_event {
        CA_EVENT_LOSS,          /* loss timeout */
        CA_EVENT_ECN_NO_CE,     /* ECT set, but not CE marked */
        CA_EVENT_ECN_IS_CE,     /* received CE marked IP packet */
-       CA_EVENT_DELAYED_ACK,   /* Delayed ack is sent */
-       CA_EVENT_NON_DELAYED_ACK,
 };
 
 /* Information about inbound ACK, passed to cong_ops->in_ack_event() */
index 9fe472f2ac950c8f3f042cca547cb8f7ce820a97..7161856bcf9c7f572943f6a78676df2aa458a5f7 100644 (file)
@@ -60,6 +60,10 @@ struct xdp_sock {
        bool zc;
        /* Protects multiple processes in the control path */
        struct mutex mutex;
+       /* Mutual exclusion of NAPI TX thread and sendmsg error paths
+        * in the SKB destructor callback.
+        */
+       spinlock_t tx_completion_lock;
        u64 rx_dropped;
 };
 
index 4ca65b56084f94526435a58a8663d58054c924f4..7363f18e65a553e12f4d1cc13844dfbf2bbe6f17 100644 (file)
@@ -226,7 +226,7 @@ enum tunable_id {
        ETHTOOL_TX_COPYBREAK,
        ETHTOOL_PFC_PREVENTION_TOUT, /* timeout in msecs */
        /*
-        * Add your fresh new tubale attribute above and remember to update
+        * Add your fresh new tunable attribute above and remember to update
         * tunable_strings[] in net/core/ethtool.c
         */
        __ETHTOOL_TUNABLE_COUNT,
index 29eb659aa77a183e36082599866fb512908d1197..e3f6ed8a7064f9276ca2b57ed5ecff3364786e9d 100644 (file)
@@ -127,6 +127,10 @@ enum {
 
 #define TCP_CM_INQ             TCP_INQ
 
+#define TCP_REPAIR_ON          1
+#define TCP_REPAIR_OFF         0
+#define TCP_REPAIR_OFF_NO_WP   -1      /* Turn off without window probes */
+
 struct tcp_repair_opt {
        __u32   opt_code;
        __u32   opt_val;
index 2d49d18b793abaf60379c4050e4148a77bae732f..e016ac3afa2422f4f4e1cdd0d8417b1e6a2f55ac 100644 (file)
@@ -991,16 +991,13 @@ static void btf_int_bits_seq_show(const struct btf *btf,
                                  void *data, u8 bits_offset,
                                  struct seq_file *m)
 {
+       u16 left_shift_bits, right_shift_bits;
        u32 int_data = btf_type_int(t);
        u16 nr_bits = BTF_INT_BITS(int_data);
        u16 total_bits_offset;
        u16 nr_copy_bytes;
        u16 nr_copy_bits;
-       u8 nr_upper_bits;
-       union {
-               u64 u64_num;
-               u8  u8_nums[8];
-       } print_num;
+       u64 print_num;
 
        total_bits_offset = bits_offset + BTF_INT_OFFSET(int_data);
        data += BITS_ROUNDDOWN_BYTES(total_bits_offset);
@@ -1008,21 +1005,20 @@ static void btf_int_bits_seq_show(const struct btf *btf,
        nr_copy_bits = nr_bits + bits_offset;
        nr_copy_bytes = BITS_ROUNDUP_BYTES(nr_copy_bits);
 
-       print_num.u64_num = 0;
-       memcpy(&print_num.u64_num, data, nr_copy_bytes);
+       print_num = 0;
+       memcpy(&print_num, data, nr_copy_bytes);
 
-       /* Ditch the higher order bits */
-       nr_upper_bits = BITS_PER_BYTE_MASKED(nr_copy_bits);
-       if (nr_upper_bits) {
-               /* We need to mask out some bits of the upper byte. */
-               u8 mask = (1 << nr_upper_bits) - 1;
+#ifdef __BIG_ENDIAN_BITFIELD
+       left_shift_bits = bits_offset;
+#else
+       left_shift_bits = BITS_PER_U64 - nr_copy_bits;
+#endif
+       right_shift_bits = BITS_PER_U64 - nr_bits;
 
-               print_num.u8_nums[nr_copy_bytes - 1] &= mask;
-       }
-
-       print_num.u64_num >>= bits_offset;
+       print_num <<= left_shift_bits;
+       print_num >>= right_shift_bits;
 
-       seq_printf(m, "0x%llx", print_num.u64_num);
+       seq_printf(m, "0x%llx", print_num);
 }
 
 static void btf_int_seq_show(const struct btf *btf, const struct btf_type *t,
index 642c97f6d1b8efeb07e0281680cd1dad26d0081b..d361fc1e3bf35fd54d485e72c2e258171e3394d5 100644 (file)
@@ -334,10 +334,15 @@ int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
 {
        struct net_device *dev = dst->dev;
        struct xdp_frame *xdpf;
+       int err;
 
        if (!dev->netdev_ops->ndo_xdp_xmit)
                return -EOPNOTSUPP;
 
+       err = xdp_ok_fwd_dev(dev, xdp->data_end - xdp->data);
+       if (unlikely(err))
+               return err;
+
        xdpf = convert_to_xdp_frame(xdp);
        if (unlikely(!xdpf))
                return -EOVERFLOW;
@@ -350,7 +355,7 @@ int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
 {
        int err;
 
-       err = __xdp_generic_ok_fwd_dev(skb, dst->dev);
+       err = xdp_ok_fwd_dev(dst->dev, skb->len);
        if (unlikely(err))
                return err;
        skb->dev = dst->dev;
index 3ca2198a6d22d9ab67c61e964811046cf9c9e512..513d9dfcf4ee136dd5e6733789a996612272376d 100644 (file)
@@ -747,13 +747,15 @@ static struct htab_elem *alloc_htab_elem(struct bpf_htab *htab, void *key,
                                 * old element will be freed immediately.
                                 * Otherwise return an error
                                 */
-                               atomic_dec(&htab->count);
-                               return ERR_PTR(-E2BIG);
+                               l_new = ERR_PTR(-E2BIG);
+                               goto dec_count;
                        }
                l_new = kmalloc_node(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN,
                                     htab->map.numa_node);
-               if (!l_new)
-                       return ERR_PTR(-ENOMEM);
+               if (!l_new) {
+                       l_new = ERR_PTR(-ENOMEM);
+                       goto dec_count;
+               }
        }
 
        memcpy(l_new->key, key, key_size);
@@ -766,7 +768,8 @@ static struct htab_elem *alloc_htab_elem(struct bpf_htab *htab, void *key,
                                                  GFP_ATOMIC | __GFP_NOWARN);
                        if (!pptr) {
                                kfree(l_new);
-                               return ERR_PTR(-ENOMEM);
+                               l_new = ERR_PTR(-ENOMEM);
+                               goto dec_count;
                        }
                }
 
@@ -780,6 +783,9 @@ static struct htab_elem *alloc_htab_elem(struct bpf_htab *htab, void *key,
 
        l_new->hash = hash;
        return l_new;
+dec_count:
+       atomic_dec(&htab->count);
+       return l_new;
 }
 
 static int check_flags(struct bpf_htab *htab, struct htab_elem *l_old,
index cf7b6a6dbd1f3bc290eb3f740e46a63ab3eba116..98fb7938beea9dd18a255ad77ebe797b01660dea 100644 (file)
@@ -312,10 +312,12 @@ static void bpf_tcp_close(struct sock *sk, long timeout)
        struct smap_psock *psock;
        struct sock *osk;
 
+       lock_sock(sk);
        rcu_read_lock();
        psock = smap_psock_sk(sk);
        if (unlikely(!psock)) {
                rcu_read_unlock();
+               release_sock(sk);
                return sk->sk_prot->close(sk, timeout);
        }
 
@@ -371,6 +373,7 @@ static void bpf_tcp_close(struct sock *sk, long timeout)
                e = psock_map_pop(sk, psock);
        }
        rcu_read_unlock();
+       release_sock(sk);
        close_fun(sk, timeout);
 }
 
@@ -568,7 +571,8 @@ static int free_sg(struct sock *sk, int start, struct sk_msg_buff *md)
        while (sg[i].length) {
                free += sg[i].length;
                sk_mem_uncharge(sk, sg[i].length);
-               put_page(sg_page(&sg[i]));
+               if (!md->skb)
+                       put_page(sg_page(&sg[i]));
                sg[i].length = 0;
                sg[i].page_link = 0;
                sg[i].offset = 0;
@@ -577,6 +581,8 @@ static int free_sg(struct sock *sk, int start, struct sk_msg_buff *md)
                if (i == MAX_SKB_FRAGS)
                        i = 0;
        }
+       if (md->skb)
+               consume_skb(md->skb);
 
        return free;
 }
@@ -1230,7 +1236,7 @@ static int smap_verdict_func(struct smap_psock *psock, struct sk_buff *skb)
         */
        TCP_SKB_CB(skb)->bpf.sk_redir = NULL;
        skb->sk = psock->sock;
-       bpf_compute_data_pointers(skb);
+       bpf_compute_data_end_sk_skb(skb);
        preempt_disable();
        rc = (*prog->bpf_func)(skb, prog->insnsi);
        preempt_enable();
@@ -1485,7 +1491,7 @@ static int smap_parse_func_strparser(struct strparser *strp,
         * any socket yet.
         */
        skb->sk = psock->sock;
-       bpf_compute_data_pointers(skb);
+       bpf_compute_data_end_sk_skb(skb);
        rc = (*prog->bpf_func)(skb, prog->insnsi);
        skb->sk = NULL;
        rcu_read_unlock();
@@ -1896,7 +1902,7 @@ static int __sock_map_ctx_update_elem(struct bpf_map *map,
                e = kzalloc(sizeof(*e), GFP_ATOMIC | __GFP_NOWARN);
                if (!e) {
                        err = -ENOMEM;
-                       goto out_progs;
+                       goto out_free;
                }
        }
 
@@ -2069,7 +2075,13 @@ static int sock_map_update_elem(struct bpf_map *map,
                return -EOPNOTSUPP;
        }
 
+       lock_sock(skops.sk);
+       preempt_disable();
+       rcu_read_lock();
        err = sock_map_ctx_update_elem(&skops, map, key, flags);
+       rcu_read_unlock();
+       preempt_enable();
+       release_sock(skops.sk);
        fput(socket->file);
        return err;
 }
@@ -2342,7 +2354,10 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops,
        if (err)
                goto err;
 
-       /* bpf_map_update_elem() can be called in_irq() */
+       /* psock is valid here because otherwise above *ctx_update_elem would
+        * have thrown an error. It is safe to skip error check.
+        */
+       psock = smap_psock_sk(sock);
        raw_spin_lock_bh(&b->lock);
        l_old = lookup_elem_raw(head, hash, key, key_size);
        if (l_old && map_flags == BPF_NOEXIST) {
@@ -2360,12 +2375,6 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops,
                goto bucket_err;
        }
 
-       psock = smap_psock_sk(sock);
-       if (unlikely(!psock)) {
-               err = -EINVAL;
-               goto bucket_err;
-       }
-
        rcu_assign_pointer(e->hash_link, l_new);
        rcu_assign_pointer(e->htab,
                           container_of(map, struct bpf_htab, map));
@@ -2388,12 +2397,10 @@ static int sock_hash_ctx_update_elem(struct bpf_sock_ops_kern *skops,
        raw_spin_unlock_bh(&b->lock);
        return 0;
 bucket_err:
+       smap_release_sock(psock, sock);
        raw_spin_unlock_bh(&b->lock);
 err:
        kfree(e);
-       psock = smap_psock_sk(sock);
-       if (psock)
-               smap_release_sock(psock, sock);
        return err;
 }
 
@@ -2415,7 +2422,13 @@ static int sock_hash_update_elem(struct bpf_map *map,
                return -EINVAL;
        }
 
+       lock_sock(skops.sk);
+       preempt_disable();
+       rcu_read_lock();
        err = sock_hash_ctx_update_elem(&skops, map, key, flags);
+       rcu_read_unlock();
+       preempt_enable();
+       release_sock(skops.sk);
        fput(socket->file);
        return err;
 }
@@ -2472,10 +2485,8 @@ struct sock  *__sock_hash_lookup_elem(struct bpf_map *map, void *key)
        b = __select_bucket(htab, hash);
        head = &b->head;
 
-       raw_spin_lock_bh(&b->lock);
        l = lookup_elem_raw(head, hash, key, key_size);
        sk = l ? l->sk : NULL;
-       raw_spin_unlock_bh(&b->lock);
        return sk;
 }
 
index d10ecd78105fa66d2ffbd6b27c42fec2a6e6b09c..a31a1ba0f8eada88e03dc4a73ad9c0305cc70198 100644 (file)
@@ -735,7 +735,9 @@ static int map_update_elem(union bpf_attr *attr)
        if (bpf_map_is_dev_bound(map)) {
                err = bpf_map_offload_update_elem(map, key, value, attr->flags);
                goto out;
-       } else if (map->map_type == BPF_MAP_TYPE_CPUMAP) {
+       } else if (map->map_type == BPF_MAP_TYPE_CPUMAP ||
+                  map->map_type == BPF_MAP_TYPE_SOCKHASH ||
+                  map->map_type == BPF_MAP_TYPE_SOCKMAP) {
                err = map->ops->map_update_elem(map, key, value, attr->flags);
                goto out;
        }
index 9e2bf834f13a21090b566862a853e7567ee03ac9..63aaac52a26553fb29529790cf0350f6dafa504b 100644 (file)
@@ -5430,6 +5430,10 @@ static int jit_subprogs(struct bpf_verifier_env *env)
                if (insn->code != (BPF_JMP | BPF_CALL) ||
                    insn->src_reg != BPF_PSEUDO_CALL)
                        continue;
+               /* Upon error here we cannot fall back to interpreter but
+                * need a hard reject of the program. Thus -EFAULT is
+                * propagated in any case.
+                */
                subprog = find_subprog(env, i + insn->imm + 1);
                if (subprog < 0) {
                        WARN_ONCE(1, "verifier bug. No program starts at insn %d\n",
@@ -5450,7 +5454,7 @@ static int jit_subprogs(struct bpf_verifier_env *env)
 
        func = kcalloc(env->subprog_cnt, sizeof(prog), GFP_KERNEL);
        if (!func)
-               return -ENOMEM;
+               goto out_undo_insn;
 
        for (i = 0; i < env->subprog_cnt; i++) {
                subprog_start = subprog_end;
@@ -5515,7 +5519,7 @@ static int jit_subprogs(struct bpf_verifier_env *env)
                tmp = bpf_int_jit_compile(func[i]);
                if (tmp != func[i] || func[i]->bpf_func != old_bpf_func) {
                        verbose(env, "JIT doesn't support bpf-to-bpf calls\n");
-                       err = -EFAULT;
+                       err = -ENOTSUPP;
                        goto out_free;
                }
                cond_resched();
@@ -5552,6 +5556,7 @@ out_free:
                if (func[i])
                        bpf_jit_free(func[i]);
        kfree(func);
+out_undo_insn:
        /* cleanup main prog to be interpreted */
        prog->jit_requested = 0;
        for (i = 0, insn = prog->insnsi; i < prog->len; i++, insn++) {
@@ -5578,6 +5583,8 @@ static int fixup_call_args(struct bpf_verifier_env *env)
                err = jit_subprogs(env);
                if (err == 0)
                        return 0;
+               if (err == -EFAULT)
+                       return err;
        }
 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
        for (i = 0; i < prog->len; i++, insn++) {
index 9427b5766134cb139ef385b27f92f6027fecceca..e5c8586cf7174cfe0526dc8fb3314676601c5e57 100644 (file)
@@ -774,7 +774,7 @@ int rhashtable_walk_start_check(struct rhashtable_iter *iter)
                                skip++;
                                if (list == iter->list) {
                                        iter->p = p;
-                                       skip = skip;
+                                       iter->skip = skip;
                                        goto found;
                                }
                        }
@@ -964,8 +964,16 @@ EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
 
 static size_t rounded_hashtable_size(const struct rhashtable_params *params)
 {
-       return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
-                  (unsigned long)params->min_size);
+       size_t retsize;
+
+       if (params->nelem_hint)
+               retsize = max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
+                             (unsigned long)params->min_size);
+       else
+               retsize = max(HASH_DEFAULT_SIZE,
+                             (unsigned long)params->min_size);
+
+       return retsize;
 }
 
 static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
@@ -1022,8 +1030,6 @@ int rhashtable_init(struct rhashtable *ht,
        struct bucket_table *tbl;
        size_t size;
 
-       size = HASH_DEFAULT_SIZE;
-
        if ((!params->key_len && !params->obj_hashfn) ||
            (params->obj_hashfn && !params->obj_cmpfn))
                return -EINVAL;
@@ -1050,8 +1056,7 @@ int rhashtable_init(struct rhashtable *ht,
 
        ht->p.min_size = max_t(u16, ht->p.min_size, HASH_MIN_SIZE);
 
-       if (params->nelem_hint)
-               size = rounded_hashtable_size(&ht->p);
+       size = rounded_hashtable_size(&ht->p);
 
        if (params->locks_mul)
                ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
@@ -1143,13 +1148,14 @@ void rhashtable_free_and_destroy(struct rhashtable *ht,
                                 void (*free_fn)(void *ptr, void *arg),
                                 void *arg)
 {
-       struct bucket_table *tbl;
+       struct bucket_table *tbl, *next_tbl;
        unsigned int i;
 
        cancel_work_sync(&ht->run_work);
 
        mutex_lock(&ht->mutex);
        tbl = rht_dereference(ht->tbl, ht);
+restart:
        if (free_fn) {
                for (i = 0; i < tbl->size; i++) {
                        struct rhash_head *pos, *next;
@@ -1166,7 +1172,12 @@ void rhashtable_free_and_destroy(struct rhashtable *ht,
                }
        }
 
+       next_tbl = rht_dereference(tbl->future_tbl, ht);
        bucket_table_free(tbl);
+       if (next_tbl) {
+               tbl = next_tbl;
+               goto restart;
+       }
        mutex_unlock(&ht->mutex);
 }
 EXPORT_SYMBOL_GPL(rhashtable_free_and_destroy);
index be09a98838252f4f0c23cec0625930cf896cd0ff..73bf6a93a3cf1141a34657bf1284893199e04db9 100644 (file)
@@ -2732,7 +2732,7 @@ static int batadv_iv_gw_dump_entry(struct sk_buff *msg, u32 portid, u32 seq,
 {
        struct batadv_neigh_ifinfo *router_ifinfo = NULL;
        struct batadv_neigh_node *router;
-       struct batadv_gw_node *curr_gw;
+       struct batadv_gw_node *curr_gw = NULL;
        int ret = 0;
        void *hdr;
 
@@ -2780,6 +2780,8 @@ static int batadv_iv_gw_dump_entry(struct sk_buff *msg, u32 portid, u32 seq,
        ret = 0;
 
 out:
+       if (curr_gw)
+               batadv_gw_node_put(curr_gw);
        if (router_ifinfo)
                batadv_neigh_ifinfo_put(router_ifinfo);
        if (router)
index ec93337ee2597738e46b87dd72724d5becf3f48e..6baec4e68898c6e992e7522d2ee8c78ce62a1b08 100644 (file)
@@ -927,7 +927,7 @@ static int batadv_v_gw_dump_entry(struct sk_buff *msg, u32 portid, u32 seq,
 {
        struct batadv_neigh_ifinfo *router_ifinfo = NULL;
        struct batadv_neigh_node *router;
-       struct batadv_gw_node *curr_gw;
+       struct batadv_gw_node *curr_gw = NULL;
        int ret = 0;
        void *hdr;
 
@@ -995,6 +995,8 @@ static int batadv_v_gw_dump_entry(struct sk_buff *msg, u32 portid, u32 seq,
        ret = 0;
 
 out:
+       if (curr_gw)
+               batadv_gw_node_put(curr_gw);
        if (router_ifinfo)
                batadv_neigh_ifinfo_put(router_ifinfo);
        if (router)
index 4229b01ac7b54008e023df0ed6546a6d541498ba..87479c60670ebfbe2ad3df17130f1289d657df7b 100644 (file)
@@ -19,6 +19,7 @@
 #include "debugfs.h"
 #include "main.h"
 
+#include <linux/dcache.h>
 #include <linux/debugfs.h>
 #include <linux/err.h>
 #include <linux/errno.h>
@@ -343,6 +344,25 @@ out:
        return -ENOMEM;
 }
 
+/**
+ * batadv_debugfs_rename_hardif() - Fix debugfs path for renamed hardif
+ * @hard_iface: hard interface which was renamed
+ */
+void batadv_debugfs_rename_hardif(struct batadv_hard_iface *hard_iface)
+{
+       const char *name = hard_iface->net_dev->name;
+       struct dentry *dir;
+       struct dentry *d;
+
+       dir = hard_iface->debug_dir;
+       if (!dir)
+               return;
+
+       d = debugfs_rename(dir->d_parent, dir, dir->d_parent, name);
+       if (!d)
+               pr_err("Can't rename debugfs dir to %s\n", name);
+}
+
 /**
  * batadv_debugfs_del_hardif() - delete the base directory for a hard interface
  *  in debugfs.
@@ -413,6 +433,26 @@ out:
        return -ENOMEM;
 }
 
+/**
+ * batadv_debugfs_rename_meshif() - Fix debugfs path for renamed softif
+ * @dev: net_device which was renamed
+ */
+void batadv_debugfs_rename_meshif(struct net_device *dev)
+{
+       struct batadv_priv *bat_priv = netdev_priv(dev);
+       const char *name = dev->name;
+       struct dentry *dir;
+       struct dentry *d;
+
+       dir = bat_priv->debug_dir;
+       if (!dir)
+               return;
+
+       d = debugfs_rename(dir->d_parent, dir, dir->d_parent, name);
+       if (!d)
+               pr_err("Can't rename debugfs dir to %s\n", name);
+}
+
 /**
  * batadv_debugfs_del_meshif() - Remove interface dependent debugfs entries
  * @dev: netdev struct of the soft interface
index 37b069698b04b369e68e4e8a31c3ac01575b0178..08a592ffbee5203ac4994fc49bf9c187c2e66f8e 100644 (file)
@@ -30,8 +30,10 @@ struct net_device;
 void batadv_debugfs_init(void);
 void batadv_debugfs_destroy(void);
 int batadv_debugfs_add_meshif(struct net_device *dev);
+void batadv_debugfs_rename_meshif(struct net_device *dev);
 void batadv_debugfs_del_meshif(struct net_device *dev);
 int batadv_debugfs_add_hardif(struct batadv_hard_iface *hard_iface);
+void batadv_debugfs_rename_hardif(struct batadv_hard_iface *hard_iface);
 void batadv_debugfs_del_hardif(struct batadv_hard_iface *hard_iface);
 
 #else
@@ -49,6 +51,10 @@ static inline int batadv_debugfs_add_meshif(struct net_device *dev)
        return 0;
 }
 
+static inline void batadv_debugfs_rename_meshif(struct net_device *dev)
+{
+}
+
 static inline void batadv_debugfs_del_meshif(struct net_device *dev)
 {
 }
@@ -59,6 +65,11 @@ int batadv_debugfs_add_hardif(struct batadv_hard_iface *hard_iface)
        return 0;
 }
 
+static inline
+void batadv_debugfs_rename_hardif(struct batadv_hard_iface *hard_iface)
+{
+}
+
 static inline
 void batadv_debugfs_del_hardif(struct batadv_hard_iface *hard_iface)
 {
index c405d15befd60bdabf9f50813c3bee446238d539..2f0d42f2f913e74cf10c0c6ce89320434994cac5 100644 (file)
@@ -989,6 +989,32 @@ void batadv_hardif_remove_interfaces(void)
        rtnl_unlock();
 }
 
+/**
+ * batadv_hard_if_event_softif() - Handle events for soft interfaces
+ * @event: NETDEV_* event to handle
+ * @net_dev: net_device which generated an event
+ *
+ * Return: NOTIFY_* result
+ */
+static int batadv_hard_if_event_softif(unsigned long event,
+                                      struct net_device *net_dev)
+{
+       struct batadv_priv *bat_priv;
+
+       switch (event) {
+       case NETDEV_REGISTER:
+               batadv_sysfs_add_meshif(net_dev);
+               bat_priv = netdev_priv(net_dev);
+               batadv_softif_create_vlan(bat_priv, BATADV_NO_FLAGS);
+               break;
+       case NETDEV_CHANGENAME:
+               batadv_debugfs_rename_meshif(net_dev);
+               break;
+       }
+
+       return NOTIFY_DONE;
+}
+
 static int batadv_hard_if_event(struct notifier_block *this,
                                unsigned long event, void *ptr)
 {
@@ -997,12 +1023,8 @@ static int batadv_hard_if_event(struct notifier_block *this,
        struct batadv_hard_iface *primary_if = NULL;
        struct batadv_priv *bat_priv;
 
-       if (batadv_softif_is_valid(net_dev) && event == NETDEV_REGISTER) {
-               batadv_sysfs_add_meshif(net_dev);
-               bat_priv = netdev_priv(net_dev);
-               batadv_softif_create_vlan(bat_priv, BATADV_NO_FLAGS);
-               return NOTIFY_DONE;
-       }
+       if (batadv_softif_is_valid(net_dev))
+               return batadv_hard_if_event_softif(event, net_dev);
 
        hard_iface = batadv_hardif_get_by_netdev(net_dev);
        if (!hard_iface && (event == NETDEV_REGISTER ||
@@ -1051,6 +1073,9 @@ static int batadv_hard_if_event(struct notifier_block *this,
                if (batadv_is_wifi_hardif(hard_iface))
                        hard_iface->num_bcasts = BATADV_NUM_BCASTS_WIRELESS;
                break;
+       case NETDEV_CHANGENAME:
+               batadv_debugfs_rename_hardif(hard_iface);
+               break;
        default:
                break;
        }
index 3986551397caa5ffb6ba7338eeb4769c8b8f99fb..12a2b7d21376721d15c6a31f3e794e4270d74b5c 100644 (file)
@@ -1705,7 +1705,9 @@ static bool batadv_tt_global_add(struct batadv_priv *bat_priv,
                ether_addr_copy(common->addr, tt_addr);
                common->vid = vid;
 
-               common->flags = flags;
+               if (!is_multicast_ether_addr(common->addr))
+                       common->flags = flags & (~BATADV_TT_SYNC_MASK);
+
                tt_global_entry->roam_at = 0;
                /* node must store current time in case of roaming. This is
                 * needed to purge this entry out on timeout (if nobody claims
@@ -1768,7 +1770,8 @@ static bool batadv_tt_global_add(struct batadv_priv *bat_priv,
                 * TT_CLIENT_TEMP, therefore they have to be copied in the
                 * client entry
                 */
-               common->flags |= flags & (~BATADV_TT_SYNC_MASK);
+               if (!is_multicast_ether_addr(common->addr))
+                       common->flags |= flags & (~BATADV_TT_SYNC_MASK);
 
                /* If there is the BATADV_TT_CLIENT_ROAM flag set, there is only
                 * one originator left in the list and we previously received a
index 68c3578343b4b4d026e9df40fda98a7850757877..22a78eedf4b1447a8f42cc442615191d66ff1b99 100644 (file)
@@ -96,6 +96,7 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
        u32 size = kattr->test.data_size_in;
        u32 repeat = kattr->test.repeat;
        u32 retval, duration;
+       int hh_len = ETH_HLEN;
        struct sk_buff *skb;
        void *data;
        int ret;
@@ -131,12 +132,22 @@ int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
        skb_reset_network_header(skb);
 
        if (is_l2)
-               __skb_push(skb, ETH_HLEN);
+               __skb_push(skb, hh_len);
        if (is_direct_pkt_access)
                bpf_compute_data_pointers(skb);
        retval = bpf_test_run(prog, skb, repeat, &duration);
-       if (!is_l2)
-               __skb_push(skb, ETH_HLEN);
+       if (!is_l2) {
+               if (skb_headroom(skb) < hh_len) {
+                       int nhead = HH_DATA_ALIGN(hh_len - skb_headroom(skb));
+
+                       if (pskb_expand_head(skb, nhead, 0, GFP_USER)) {
+                               kfree_skb(skb);
+                               return -ENOMEM;
+                       }
+               }
+               memset(__skb_push(skb, hh_len), 0, hh_len);
+       }
+
        size = skb->len;
        /* bpf program can never convert linear skb to non-linear */
        if (WARN_ON_ONCE(skb_is_nonlinear(skb)))
index 0ca6907d7efee75b68450f66c2c5a7172272895e..06da770f543fdc2742f8503a5436a5893a566914 100644 (file)
@@ -459,11 +459,21 @@ static bool convert_bpf_ld_abs(struct sock_filter *fp, struct bpf_insn **insnp)
             (!unaligned_ok && offset >= 0 &&
              offset + ip_align >= 0 &&
              offset + ip_align % size == 0))) {
+               bool ldx_off_ok = offset <= S16_MAX;
+
                *insn++ = BPF_MOV64_REG(BPF_REG_TMP, BPF_REG_H);
                *insn++ = BPF_ALU64_IMM(BPF_SUB, BPF_REG_TMP, offset);
-               *insn++ = BPF_JMP_IMM(BPF_JSLT, BPF_REG_TMP, size, 2 + endian);
-               *insn++ = BPF_LDX_MEM(BPF_SIZE(fp->code), BPF_REG_A, BPF_REG_D,
-                                     offset);
+               *insn++ = BPF_JMP_IMM(BPF_JSLT, BPF_REG_TMP,
+                                     size, 2 + endian + (!ldx_off_ok * 2));
+               if (ldx_off_ok) {
+                       *insn++ = BPF_LDX_MEM(BPF_SIZE(fp->code), BPF_REG_A,
+                                             BPF_REG_D, offset);
+               } else {
+                       *insn++ = BPF_MOV64_REG(BPF_REG_TMP, BPF_REG_D);
+                       *insn++ = BPF_ALU64_IMM(BPF_ADD, BPF_REG_TMP, offset);
+                       *insn++ = BPF_LDX_MEM(BPF_SIZE(fp->code), BPF_REG_A,
+                                             BPF_REG_TMP, 0);
+               }
                if (endian)
                        *insn++ = BPF_ENDIAN(BPF_FROM_BE, BPF_REG_A, size * 8);
                *insn++ = BPF_JMP_A(8);
@@ -1762,6 +1772,37 @@ static const struct bpf_func_proto bpf_skb_pull_data_proto = {
        .arg2_type      = ARG_ANYTHING,
 };
 
+static inline int sk_skb_try_make_writable(struct sk_buff *skb,
+                                          unsigned int write_len)
+{
+       int err = __bpf_try_make_writable(skb, write_len);
+
+       bpf_compute_data_end_sk_skb(skb);
+       return err;
+}
+
+BPF_CALL_2(sk_skb_pull_data, struct sk_buff *, skb, u32, len)
+{
+       /* Idea is the following: should the needed direct read/write
+        * test fail during runtime, we can pull in more data and redo
+        * again, since implicitly, we invalidate previous checks here.
+        *
+        * Or, since we know how much we need to make read/writeable,
+        * this can be done once at the program beginning for direct
+        * access case. By this we overcome limitations of only current
+        * headroom being accessible.
+        */
+       return sk_skb_try_make_writable(skb, len ? : skb_headlen(skb));
+}
+
+static const struct bpf_func_proto sk_skb_pull_data_proto = {
+       .func           = sk_skb_pull_data,
+       .gpl_only       = false,
+       .ret_type       = RET_INTEGER,
+       .arg1_type      = ARG_PTR_TO_CTX,
+       .arg2_type      = ARG_ANYTHING,
+};
+
 BPF_CALL_5(bpf_l3_csum_replace, struct sk_buff *, skb, u32, offset,
           u64, from, u64, to, u64, flags)
 {
@@ -2779,7 +2820,8 @@ static int bpf_skb_net_shrink(struct sk_buff *skb, u32 len_diff)
 
 static u32 __bpf_skb_max_len(const struct sk_buff *skb)
 {
-       return skb->dev->mtu + skb->dev->hard_header_len;
+       return skb->dev ? skb->dev->mtu + skb->dev->hard_header_len :
+                         SKB_MAX_ALLOC;
 }
 
 static int bpf_skb_adjust_net(struct sk_buff *skb, s32 len_diff)
@@ -2863,8 +2905,8 @@ static int bpf_skb_trim_rcsum(struct sk_buff *skb, unsigned int new_len)
        return __skb_trim_rcsum(skb, new_len);
 }
 
-BPF_CALL_3(bpf_skb_change_tail, struct sk_buff *, skb, u32, new_len,
-          u64, flags)
+static inline int __bpf_skb_change_tail(struct sk_buff *skb, u32 new_len,
+                                       u64 flags)
 {
        u32 max_len = __bpf_skb_max_len(skb);
        u32 min_len = __bpf_skb_min_len(skb);
@@ -2900,6 +2942,13 @@ BPF_CALL_3(bpf_skb_change_tail, struct sk_buff *, skb, u32, new_len,
                if (!ret && skb_is_gso(skb))
                        skb_gso_reset(skb);
        }
+       return ret;
+}
+
+BPF_CALL_3(bpf_skb_change_tail, struct sk_buff *, skb, u32, new_len,
+          u64, flags)
+{
+       int ret = __bpf_skb_change_tail(skb, new_len, flags);
 
        bpf_compute_data_pointers(skb);
        return ret;
@@ -2914,8 +2963,26 @@ static const struct bpf_func_proto bpf_skb_change_tail_proto = {
        .arg3_type      = ARG_ANYTHING,
 };
 
-BPF_CALL_3(bpf_skb_change_head, struct sk_buff *, skb, u32, head_room,
+BPF_CALL_3(sk_skb_change_tail, struct sk_buff *, skb, u32, new_len,
           u64, flags)
+{
+       int ret = __bpf_skb_change_tail(skb, new_len, flags);
+
+       bpf_compute_data_end_sk_skb(skb);
+       return ret;
+}
+
+static const struct bpf_func_proto sk_skb_change_tail_proto = {
+       .func           = sk_skb_change_tail,
+       .gpl_only       = false,
+       .ret_type       = RET_INTEGER,
+       .arg1_type      = ARG_PTR_TO_CTX,
+       .arg2_type      = ARG_ANYTHING,
+       .arg3_type      = ARG_ANYTHING,
+};
+
+static inline int __bpf_skb_change_head(struct sk_buff *skb, u32 head_room,
+                                       u64 flags)
 {
        u32 max_len = __bpf_skb_max_len(skb);
        u32 new_len = skb->len + head_room;
@@ -2941,8 +3008,16 @@ BPF_CALL_3(bpf_skb_change_head, struct sk_buff *, skb, u32, head_room,
                skb_reset_mac_header(skb);
        }
 
+       return ret;
+}
+
+BPF_CALL_3(bpf_skb_change_head, struct sk_buff *, skb, u32, head_room,
+          u64, flags)
+{
+       int ret = __bpf_skb_change_head(skb, head_room, flags);
+
        bpf_compute_data_pointers(skb);
-       return 0;
+       return ret;
 }
 
 static const struct bpf_func_proto bpf_skb_change_head_proto = {
@@ -2954,6 +3029,23 @@ static const struct bpf_func_proto bpf_skb_change_head_proto = {
        .arg3_type      = ARG_ANYTHING,
 };
 
+BPF_CALL_3(sk_skb_change_head, struct sk_buff *, skb, u32, head_room,
+          u64, flags)
+{
+       int ret = __bpf_skb_change_head(skb, head_room, flags);
+
+       bpf_compute_data_end_sk_skb(skb);
+       return ret;
+}
+
+static const struct bpf_func_proto sk_skb_change_head_proto = {
+       .func           = sk_skb_change_head,
+       .gpl_only       = false,
+       .ret_type       = RET_INTEGER,
+       .arg1_type      = ARG_PTR_TO_CTX,
+       .arg2_type      = ARG_ANYTHING,
+       .arg3_type      = ARG_ANYTHING,
+};
 static unsigned long xdp_get_metalen(const struct xdp_buff *xdp)
 {
        return xdp_data_meta_unsupported(xdp) ? 0 :
@@ -3046,12 +3138,16 @@ static int __bpf_tx_xdp(struct net_device *dev,
                        u32 index)
 {
        struct xdp_frame *xdpf;
-       int sent;
+       int err, sent;
 
        if (!dev->netdev_ops->ndo_xdp_xmit) {
                return -EOPNOTSUPP;
        }
 
+       err = xdp_ok_fwd_dev(dev, xdp->data_end - xdp->data);
+       if (unlikely(err))
+               return err;
+
        xdpf = convert_to_xdp_frame(xdp);
        if (unlikely(!xdpf))
                return -EOVERFLOW;
@@ -3285,7 +3381,8 @@ int xdp_do_generic_redirect(struct net_device *dev, struct sk_buff *skb,
                goto err;
        }
 
-       if (unlikely((err = __xdp_generic_ok_fwd_dev(skb, fwd))))
+       err = xdp_ok_fwd_dev(fwd, skb->len);
+       if (unlikely(err))
                goto err;
 
        skb->dev = fwd;
@@ -4439,10 +4536,10 @@ static const struct bpf_func_proto bpf_lwt_push_encap_proto = {
        .arg4_type      = ARG_CONST_SIZE
 };
 
+#if IS_ENABLED(CONFIG_IPV6_SEG6_BPF)
 BPF_CALL_4(bpf_lwt_seg6_store_bytes, struct sk_buff *, skb, u32, offset,
           const void *, from, u32, len)
 {
-#if IS_ENABLED(CONFIG_IPV6_SEG6_BPF)
        struct seg6_bpf_srh_state *srh_state =
                this_cpu_ptr(&seg6_bpf_srh_states);
        void *srh_tlvs, *srh_end, *ptr;
@@ -4468,9 +4565,6 @@ BPF_CALL_4(bpf_lwt_seg6_store_bytes, struct sk_buff *, skb, u32, offset,
 
        memcpy(skb->data + offset, from, len);
        return 0;
-#else /* CONFIG_IPV6_SEG6_BPF */
-       return -EOPNOTSUPP;
-#endif
 }
 
 static const struct bpf_func_proto bpf_lwt_seg6_store_bytes_proto = {
@@ -4486,7 +4580,6 @@ static const struct bpf_func_proto bpf_lwt_seg6_store_bytes_proto = {
 BPF_CALL_4(bpf_lwt_seg6_action, struct sk_buff *, skb,
           u32, action, void *, param, u32, param_len)
 {
-#if IS_ENABLED(CONFIG_IPV6_SEG6_BPF)
        struct seg6_bpf_srh_state *srh_state =
                this_cpu_ptr(&seg6_bpf_srh_states);
        struct ipv6_sr_hdr *srh;
@@ -4534,9 +4627,6 @@ BPF_CALL_4(bpf_lwt_seg6_action, struct sk_buff *, skb,
        default:
                return -EINVAL;
        }
-#else /* CONFIG_IPV6_SEG6_BPF */
-       return -EOPNOTSUPP;
-#endif
 }
 
 static const struct bpf_func_proto bpf_lwt_seg6_action_proto = {
@@ -4552,7 +4642,6 @@ static const struct bpf_func_proto bpf_lwt_seg6_action_proto = {
 BPF_CALL_3(bpf_lwt_seg6_adjust_srh, struct sk_buff *, skb, u32, offset,
           s32, len)
 {
-#if IS_ENABLED(CONFIG_IPV6_SEG6_BPF)
        struct seg6_bpf_srh_state *srh_state =
                this_cpu_ptr(&seg6_bpf_srh_states);
        void *srh_end, *srh_tlvs, *ptr;
@@ -4596,9 +4685,6 @@ BPF_CALL_3(bpf_lwt_seg6_adjust_srh, struct sk_buff *, skb, u32, offset,
        srh_state->hdrlen += len;
        srh_state->valid = 0;
        return 0;
-#else /* CONFIG_IPV6_SEG6_BPF */
-       return -EOPNOTSUPP;
-#endif
 }
 
 static const struct bpf_func_proto bpf_lwt_seg6_adjust_srh_proto = {
@@ -4609,6 +4695,7 @@ static const struct bpf_func_proto bpf_lwt_seg6_adjust_srh_proto = {
        .arg2_type      = ARG_ANYTHING,
        .arg3_type      = ARG_ANYTHING,
 };
+#endif /* CONFIG_IPV6_SEG6_BPF */
 
 bool bpf_helper_changes_pkt_data(void *func)
 {
@@ -4617,9 +4704,12 @@ bool bpf_helper_changes_pkt_data(void *func)
            func == bpf_skb_store_bytes ||
            func == bpf_skb_change_proto ||
            func == bpf_skb_change_head ||
+           func == sk_skb_change_head ||
            func == bpf_skb_change_tail ||
+           func == sk_skb_change_tail ||
            func == bpf_skb_adjust_room ||
            func == bpf_skb_pull_data ||
+           func == sk_skb_pull_data ||
            func == bpf_clone_redirect ||
            func == bpf_l3_csum_replace ||
            func == bpf_l4_csum_replace ||
@@ -4627,11 +4717,12 @@ bool bpf_helper_changes_pkt_data(void *func)
            func == bpf_xdp_adjust_meta ||
            func == bpf_msg_pull_data ||
            func == bpf_xdp_adjust_tail ||
-           func == bpf_lwt_push_encap ||
+#if IS_ENABLED(CONFIG_IPV6_SEG6_BPF)
            func == bpf_lwt_seg6_store_bytes ||
            func == bpf_lwt_seg6_adjust_srh ||
-           func == bpf_lwt_seg6_action
-           )
+           func == bpf_lwt_seg6_action ||
+#endif
+           func == bpf_lwt_push_encap)
                return true;
 
        return false;
@@ -4871,11 +4962,11 @@ sk_skb_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
        case BPF_FUNC_skb_load_bytes:
                return &bpf_skb_load_bytes_proto;
        case BPF_FUNC_skb_pull_data:
-               return &bpf_skb_pull_data_proto;
+               return &sk_skb_pull_data_proto;
        case BPF_FUNC_skb_change_tail:
-               return &bpf_skb_change_tail_proto;
+               return &sk_skb_change_tail_proto;
        case BPF_FUNC_skb_change_head:
-               return &bpf_skb_change_head_proto;
+               return &sk_skb_change_head_proto;
        case BPF_FUNC_get_socket_cookie:
                return &bpf_get_socket_cookie_proto;
        case BPF_FUNC_get_socket_uid:
@@ -4966,12 +5057,14 @@ static const struct bpf_func_proto *
 lwt_seg6local_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog)
 {
        switch (func_id) {
+#if IS_ENABLED(CONFIG_IPV6_SEG6_BPF)
        case BPF_FUNC_lwt_seg6_store_bytes:
                return &bpf_lwt_seg6_store_bytes_proto;
        case BPF_FUNC_lwt_seg6_action:
                return &bpf_lwt_seg6_action_proto;
        case BPF_FUNC_lwt_seg6_adjust_srh:
                return &bpf_lwt_seg6_adjust_srh_proto;
+#endif
        default:
                return lwt_out_func_proto(func_id, prog);
        }
index b2b2323bdc84c44afc33304d9d2f6a22738f6523..188d693cb251a05d6483b81bbd8d815e28b77164 100644 (file)
@@ -77,8 +77,20 @@ gnet_stats_start_copy_compat(struct sk_buff *skb, int type, int tc_stats_type,
                d->lock = lock;
                spin_lock_bh(lock);
        }
-       if (d->tail)
-               return gnet_stats_copy(d, type, NULL, 0, padattr);
+       if (d->tail) {
+               int ret = gnet_stats_copy(d, type, NULL, 0, padattr);
+
+               /* The initial attribute added in gnet_stats_copy() may be
+                * preceded by a padding attribute, in which case d->tail will
+                * end up pointing at the padding instead of the real attribute.
+                * Fix this so gnet_stats_finish_copy() adjusts the length of
+                * the right attribute.
+                */
+               if (ret == 0 && d->tail->nla_type == padattr)
+                       d->tail = (struct nlattr *)((char *)d->tail +
+                                                   NLA_ALIGN(d->tail->nla_len));
+               return ret;
+       }
 
        return 0;
 }
index eba8dae22c25b2fe0e49c65392f9c69760f1396f..8e51f8555e11b95bc48ab334f50571048f705101 100644 (file)
@@ -858,6 +858,7 @@ static struct sk_buff *__skb_clone(struct sk_buff *n, struct sk_buff *skb)
        n->cloned = 1;
        n->nohdr = 0;
        n->peeked = 0;
+       C(pfmemalloc);
        n->destructor = NULL;
        C(tail);
        C(end);
index 40c851693f77e35a1f573fdbf0bcd86adb94cf13..0c9478b91fa5b6c8f6b586ed8ead66c8db538ea7 100644 (file)
@@ -86,35 +86,39 @@ dns_resolver_preparse(struct key_preparsed_payload *prep)
                opt++;
                kdebug("options: '%s'", opt);
                do {
+                       int opt_len, opt_nlen;
                        const char *eq;
-                       int opt_len, opt_nlen, opt_vlen, tmp;
+                       char optval[128];
 
                        next_opt = memchr(opt, '#', end - opt) ?: end;
                        opt_len = next_opt - opt;
-                       if (opt_len <= 0 || opt_len > 128) {
+                       if (opt_len <= 0 || opt_len > sizeof(optval)) {
                                pr_warn_ratelimited("Invalid option length (%d) for dns_resolver key\n",
                                                    opt_len);
                                return -EINVAL;
                        }
 
-                       eq = memchr(opt, '=', opt_len) ?: end;
-                       opt_nlen = eq - opt;
-                       eq++;
-                       opt_vlen = next_opt - eq; /* will be -1 if no value */
+                       eq = memchr(opt, '=', opt_len);
+                       if (eq) {
+                               opt_nlen = eq - opt;
+                               eq++;
+                               memcpy(optval, eq, next_opt - eq);
+                               optval[next_opt - eq] = '\0';
+                       } else {
+                               opt_nlen = opt_len;
+                               optval[0] = '\0';
+                       }
 
-                       tmp = opt_vlen >= 0 ? opt_vlen : 0;
-                       kdebug("option '%*.*s' val '%*.*s'",
-                              opt_nlen, opt_nlen, opt, tmp, tmp, eq);
+                       kdebug("option '%*.*s' val '%s'",
+                              opt_nlen, opt_nlen, opt, optval);
 
                        /* see if it's an error number representing a DNS error
                         * that's to be recorded as the result in this key */
                        if (opt_nlen == sizeof(DNS_ERRORNO_OPTION) - 1 &&
                            memcmp(opt, DNS_ERRORNO_OPTION, opt_nlen) == 0) {
                                kdebug("dns error number option");
-                               if (opt_vlen <= 0)
-                                       goto bad_option_value;
 
-                               ret = kstrtoul(eq, 10, &derrno);
+                               ret = kstrtoul(optval, 10, &derrno);
                                if (ret < 0)
                                        goto bad_option_value;
 
index 275449b0d633586a4befec517ab3a36c5e3ba5a5..3297e7fa99458b13c40609588f187d366cf37411 100644 (file)
@@ -90,12 +90,18 @@ static int lowpan_neigh_construct(struct net_device *dev, struct neighbour *n)
        return 0;
 }
 
+static int lowpan_get_iflink(const struct net_device *dev)
+{
+       return lowpan_802154_dev(dev)->wdev->ifindex;
+}
+
 static const struct net_device_ops lowpan_netdev_ops = {
        .ndo_init               = lowpan_dev_init,
        .ndo_start_xmit         = lowpan_xmit,
        .ndo_open               = lowpan_open,
        .ndo_stop               = lowpan_stop,
        .ndo_neigh_construct    = lowpan_neigh_construct,
+       .ndo_get_iflink         = lowpan_get_iflink,
 };
 
 static void lowpan_setup(struct net_device *ldev)
index b21833651394233bbdb143d765e4408333b13b72..e46cdd310e5f86ef6985993e4226db614a2a8732 100644 (file)
@@ -300,6 +300,7 @@ __be32 fib_compute_spec_dst(struct sk_buff *skb)
        if (!ipv4_is_zeronet(ip_hdr(skb)->saddr)) {
                struct flowi4 fl4 = {
                        .flowi4_iif = LOOPBACK_IFINDEX,
+                       .flowi4_oif = l3mdev_master_ifindex_rcu(dev),
                        .daddr = ip_hdr(skb)->saddr,
                        .flowi4_tos = RT_TOS(ip_hdr(skb)->tos),
                        .flowi4_scope = scope,
index 85b617b655bc2d602563b1bd174f436554c9d046..b3c899a630a0fe696207f8749f5ceb510daa445e 100644 (file)
@@ -1200,13 +1200,14 @@ static void igmpv3_del_delrec(struct in_device *in_dev, struct ip_mc_list *im)
        spin_lock_bh(&im->lock);
        if (pmc) {
                im->interface = pmc->interface;
-               im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
                im->sfmode = pmc->sfmode;
                if (pmc->sfmode == MCAST_INCLUDE) {
                        im->tomb = pmc->tomb;
                        im->sources = pmc->sources;
                        for (psf = im->sources; psf; psf = psf->sf_next)
-                               psf->sf_crcount = im->crcount;
+                               psf->sf_crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
+               } else {
+                       im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
                }
                in_dev_put(pmc->interface);
                kfree(pmc);
@@ -1288,7 +1289,7 @@ static void igmp_group_dropped(struct ip_mc_list *im)
 #endif
 }
 
-static void igmp_group_added(struct ip_mc_list *im)
+static void igmp_group_added(struct ip_mc_list *im, unsigned int mode)
 {
        struct in_device *in_dev = im->interface;
 #ifdef CONFIG_IP_MULTICAST
@@ -1316,7 +1317,13 @@ static void igmp_group_added(struct ip_mc_list *im)
        }
        /* else, v3 */
 
-       im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
+       /* Based on RFC3376 5.1, for newly added INCLUDE SSM, we should
+        * not send filter-mode change record as the mode should be from
+        * IN() to IN(A).
+        */
+       if (mode == MCAST_EXCLUDE)
+               im->crcount = in_dev->mr_qrv ?: net->ipv4.sysctl_igmp_qrv;
+
        igmp_ifc_event(in_dev);
 #endif
 }
@@ -1381,8 +1388,7 @@ static void ip_mc_hash_remove(struct in_device *in_dev,
 /*
  *     A socket has joined a multicast group on device dev.
  */
-
-void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
+void __ip_mc_inc_group(struct in_device *in_dev, __be32 addr, unsigned int mode)
 {
        struct ip_mc_list *im;
 #ifdef CONFIG_IP_MULTICAST
@@ -1394,7 +1400,7 @@ void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
        for_each_pmc_rtnl(in_dev, im) {
                if (im->multiaddr == addr) {
                        im->users++;
-                       ip_mc_add_src(in_dev, &addr, MCAST_EXCLUDE, 0, NULL, 0);
+                       ip_mc_add_src(in_dev, &addr, mode, 0, NULL, 0);
                        goto out;
                }
        }
@@ -1408,8 +1414,8 @@ void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
        in_dev_hold(in_dev);
        im->multiaddr = addr;
        /* initial mode is (EX, empty) */
-       im->sfmode = MCAST_EXCLUDE;
-       im->sfcount[MCAST_EXCLUDE] = 1;
+       im->sfmode = mode;
+       im->sfcount[mode] = 1;
        refcount_set(&im->refcnt, 1);
        spin_lock_init(&im->lock);
 #ifdef CONFIG_IP_MULTICAST
@@ -1426,12 +1432,17 @@ void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
 #ifdef CONFIG_IP_MULTICAST
        igmpv3_del_delrec(in_dev, im);
 #endif
-       igmp_group_added(im);
+       igmp_group_added(im, mode);
        if (!in_dev->dead)
                ip_rt_multicast_event(in_dev);
 out:
        return;
 }
+
+void ip_mc_inc_group(struct in_device *in_dev, __be32 addr)
+{
+       __ip_mc_inc_group(in_dev, addr, MCAST_EXCLUDE);
+}
 EXPORT_SYMBOL(ip_mc_inc_group);
 
 static int ip_mc_check_iphdr(struct sk_buff *skb)
@@ -1688,7 +1699,7 @@ void ip_mc_remap(struct in_device *in_dev)
 #ifdef CONFIG_IP_MULTICAST
                igmpv3_del_delrec(in_dev, pmc);
 #endif
-               igmp_group_added(pmc);
+               igmp_group_added(pmc, pmc->sfmode);
        }
 }
 
@@ -1751,7 +1762,7 @@ void ip_mc_up(struct in_device *in_dev)
 #ifdef CONFIG_IP_MULTICAST
                igmpv3_del_delrec(in_dev, pmc);
 #endif
-               igmp_group_added(pmc);
+               igmp_group_added(pmc, pmc->sfmode);
        }
 }
 
@@ -2130,8 +2141,8 @@ static void ip_mc_clear_src(struct ip_mc_list *pmc)
 
 /* Join a multicast group
  */
-
-int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
+static int __ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr,
+                             unsigned int mode)
 {
        __be32 addr = imr->imr_multiaddr.s_addr;
        struct ip_mc_socklist *iml, *i;
@@ -2172,15 +2183,30 @@ int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
        memcpy(&iml->multi, imr, sizeof(*imr));
        iml->next_rcu = inet->mc_list;
        iml->sflist = NULL;
-       iml->sfmode = MCAST_EXCLUDE;
+       iml->sfmode = mode;
        rcu_assign_pointer(inet->mc_list, iml);
-       ip_mc_inc_group(in_dev, addr);
+       __ip_mc_inc_group(in_dev, addr, mode);
        err = 0;
 done:
        return err;
 }
+
+/* Join ASM (Any-Source Multicast) group
+ */
+int ip_mc_join_group(struct sock *sk, struct ip_mreqn *imr)
+{
+       return __ip_mc_join_group(sk, imr, MCAST_EXCLUDE);
+}
 EXPORT_SYMBOL(ip_mc_join_group);
 
+/* Join SSM (Source-Specific Multicast) group
+ */
+int ip_mc_join_group_ssm(struct sock *sk, struct ip_mreqn *imr,
+                        unsigned int mode)
+{
+       return __ip_mc_join_group(sk, imr, mode);
+}
+
 static int ip_mc_leave_src(struct sock *sk, struct ip_mc_socklist *iml,
                           struct in_device *in_dev)
 {
index c9e35b81d0931df8429a33e8d03e719b87da0747..1e4cf3ab560fac154fefb7acd3539eb6e91ed84e 100644 (file)
@@ -90,7 +90,7 @@ static void inet_frags_free_cb(void *ptr, void *arg)
 
 void inet_frags_exit_net(struct netns_frags *nf)
 {
-       nf->low_thresh = 0; /* prevent creation of new frags */
+       nf->high_thresh = 0; /* prevent creation of new frags */
 
        rhashtable_free_and_destroy(&nf->rhashtable, inet_frags_free_cb, NULL);
 }
index fc32fdbeefa61c18da5b9330d4da73ca6db992bd..64c76dcf73863a81addc1c7effcd3a687a97483d 100644 (file)
@@ -984,7 +984,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
                        mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr;
                        mreq.imr_address.s_addr = mreqs.imr_interface;
                        mreq.imr_ifindex = 0;
-                       err = ip_mc_join_group(sk, &mreq);
+                       err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
                        if (err && err != -EADDRINUSE)
                                break;
                        omode = MCAST_INCLUDE;
@@ -1061,7 +1061,7 @@ static int do_ip_setsockopt(struct sock *sk, int level,
                        mreq.imr_multiaddr = psin->sin_addr;
                        mreq.imr_address.s_addr = 0;
                        mreq.imr_ifindex = greqs.gsr_interface;
-                       err = ip_mc_join_group(sk, &mreq);
+                       err = ip_mc_join_group_ssm(sk, &mreq, MCAST_INCLUDE);
                        if (err && err != -EADDRINUSE)
                                break;
                        greqs.gsr_interface = mreq.imr_ifindex;
index ca0dad90803a92bdcbb1e199554985ad4626fada..e77872c93c206693f4bcfdde98a044c6e7cfb780 100644 (file)
@@ -1898,6 +1898,7 @@ static struct xt_match ipt_builtin_mt[] __read_mostly = {
                .checkentry = icmp_checkentry,
                .proto      = IPPROTO_ICMP,
                .family     = NFPROTO_IPV4,
+               .me         = THIS_MODULE,
        },
 };
 
index 805e83ec3ad9347abc6ce778f296319746772f1c..16471410496592f52ac7927d218a44341f139339 100644 (file)
@@ -37,7 +37,7 @@ nf_tproxy_handle_time_wait4(struct net *net, struct sk_buff *skb,
                 * to a listener socket if there's one */
                struct sock *sk2;
 
-               sk2 = nf_tproxy_get_sock_v4(net, skb, hp, iph->protocol,
+               sk2 = nf_tproxy_get_sock_v4(net, skb, iph->protocol,
                                            iph->saddr, laddr ? laddr : iph->daddr,
                                            hp->source, lport ? lport : hp->dest,
                                            skb->dev, NF_TPROXY_LOOKUP_LISTENER);
@@ -71,7 +71,7 @@ __be32 nf_tproxy_laddr4(struct sk_buff *skb, __be32 user_laddr, __be32 daddr)
 EXPORT_SYMBOL_GPL(nf_tproxy_laddr4);
 
 struct sock *
-nf_tproxy_get_sock_v4(struct net *net, struct sk_buff *skb, void *hp,
+nf_tproxy_get_sock_v4(struct net *net, struct sk_buff *skb,
                      const u8 protocol,
                      const __be32 saddr, const __be32 daddr,
                      const __be16 sport, const __be16 dport,
@@ -79,16 +79,21 @@ nf_tproxy_get_sock_v4(struct net *net, struct sk_buff *skb, void *hp,
                      const enum nf_tproxy_lookup_t lookup_type)
 {
        struct sock *sk;
-       struct tcphdr *tcph;
 
        switch (protocol) {
-       case IPPROTO_TCP:
+       case IPPROTO_TCP: {
+               struct tcphdr _hdr, *hp;
+
+               hp = skb_header_pointer(skb, ip_hdrlen(skb),
+                                       sizeof(struct tcphdr), &_hdr);
+               if (hp == NULL)
+                       return NULL;
+
                switch (lookup_type) {
                case NF_TPROXY_LOOKUP_LISTENER:
-                       tcph = hp;
                        sk = inet_lookup_listener(net, &tcp_hashinfo, skb,
                                                    ip_hdrlen(skb) +
-                                                     __tcp_hdrlen(tcph),
+                                                     __tcp_hdrlen(hp),
                                                    saddr, sport,
                                                    daddr, dport,
                                                    in->ifindex, 0);
@@ -110,6 +115,7 @@ nf_tproxy_get_sock_v4(struct net *net, struct sk_buff *skb, void *hp,
                        BUG();
                }
                break;
+               }
        case IPPROTO_UDP:
                sk = udp4_lib_lookup(net, saddr, sport, daddr, dport,
                                     in->ifindex);
index af0a857d8352f2bc94d4f6baa0438a7ac9b2157b..5fa335fd385254def583b9a5100fbe7b9ce94cd6 100644 (file)
@@ -189,8 +189,9 @@ static int ipv4_ping_group_range(struct ctl_table *table, int write,
        if (write && ret == 0) {
                low = make_kgid(user_ns, urange[0]);
                high = make_kgid(user_ns, urange[1]);
-               if (!gid_valid(low) || !gid_valid(high) ||
-                   (urange[1] < urange[0]) || gid_lt(high, low)) {
+               if (!gid_valid(low) || !gid_valid(high))
+                       return -EINVAL;
+               if (urange[1] < urange[0] || gid_lt(high, low)) {
                        low = make_kgid(&init_user_ns, 1);
                        high = make_kgid(&init_user_ns, 0);
                }
index e7b53d2a971f266bfa8d6403063fe82de61aaebe..4491faf83f4f93cf4384f7b192ffe3022567cc0a 100644 (file)
@@ -1998,7 +1998,7 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
                         * shouldn't happen.
                         */
                        if (WARN(before(*seq, TCP_SKB_CB(skb)->seq),
-                                "recvmsg bug: copied %X seq %X rcvnxt %X fl %X\n",
+                                "TCP recvmsg seq # bug: copied %X, seq %X, rcvnxt %X, fl %X\n",
                                 *seq, TCP_SKB_CB(skb)->seq, tp->rcv_nxt,
                                 flags))
                                break;
@@ -2013,7 +2013,7 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
                        if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
                                goto found_fin_ok;
                        WARN(!(flags & MSG_PEEK),
-                            "recvmsg bug 2: copied %X seq %X rcvnxt %X fl %X\n",
+                            "TCP recvmsg seq # bug 2: copied %X, seq %X, rcvnxt %X, fl %X\n",
                             *seq, TCP_SKB_CB(skb)->seq, tp->rcv_nxt, flags);
                }
 
@@ -2562,6 +2562,8 @@ int tcp_disconnect(struct sock *sk, int flags)
 
        tcp_clear_xmit_timers(sk);
        __skb_queue_purge(&sk->sk_receive_queue);
+       tp->copied_seq = tp->rcv_nxt;
+       tp->urg_data = 0;
        tcp_write_queue_purge(sk);
        tcp_fastopen_active_disable_ofo_check(sk);
        skb_rbtree_purge(&tp->out_of_order_queue);
@@ -2821,14 +2823,17 @@ static int do_tcp_setsockopt(struct sock *sk, int level,
        case TCP_REPAIR:
                if (!tcp_can_repair_sock(sk))
                        err = -EPERM;
-               else if (val == 1) {
+               else if (val == TCP_REPAIR_ON) {
                        tp->repair = 1;
                        sk->sk_reuse = SK_FORCE_REUSE;
                        tp->repair_queue = TCP_NO_QUEUE;
-               } else if (val == 0) {
+               } else if (val == TCP_REPAIR_OFF) {
                        tp->repair = 0;
                        sk->sk_reuse = SK_NO_REUSE;
                        tcp_send_window_probe(sk);
+               } else if (val == TCP_REPAIR_OFF_NO_WP) {
+                       tp->repair = 0;
+                       sk->sk_reuse = SK_NO_REUSE;
                } else
                        err = -EINVAL;
 
@@ -3720,8 +3725,7 @@ int tcp_abort(struct sock *sk, int err)
                        struct request_sock *req = inet_reqsk(sk);
 
                        local_bh_disable();
-                       inet_csk_reqsk_queue_drop_and_put(req->rsk_listener,
-                                                         req);
+                       inet_csk_reqsk_queue_drop(req->rsk_listener, req);
                        local_bh_enable();
                        return 0;
                }
index 5f5e5936760e65739859d0d8d9717b3204482a43..5869f89ca6564ec46a3c04a2b344a8630b0ba459 100644 (file)
@@ -55,7 +55,6 @@ struct dctcp {
        u32 dctcp_alpha;
        u32 next_seq;
        u32 ce_state;
-       u32 delayed_ack_reserved;
        u32 loss_cwnd;
 };
 
@@ -96,7 +95,6 @@ static void dctcp_init(struct sock *sk)
 
                ca->dctcp_alpha = min(dctcp_alpha_on_init, DCTCP_MAX_ALPHA);
 
-               ca->delayed_ack_reserved = 0;
                ca->loss_cwnd = 0;
                ca->ce_state = 0;
 
@@ -134,7 +132,8 @@ static void dctcp_ce_state_0_to_1(struct sock *sk)
        /* State has changed from CE=0 to CE=1 and delayed
         * ACK has not sent yet.
         */
-       if (!ca->ce_state && ca->delayed_ack_reserved) {
+       if (!ca->ce_state &&
+           inet_csk(sk)->icsk_ack.pending & ICSK_ACK_TIMER) {
                u32 tmp_rcv_nxt;
 
                /* Save current rcv_nxt. */
@@ -164,7 +163,8 @@ static void dctcp_ce_state_1_to_0(struct sock *sk)
        /* State has changed from CE=1 to CE=0 and delayed
         * ACK has not sent yet.
         */
-       if (ca->ce_state && ca->delayed_ack_reserved) {
+       if (ca->ce_state &&
+           inet_csk(sk)->icsk_ack.pending & ICSK_ACK_TIMER) {
                u32 tmp_rcv_nxt;
 
                /* Save current rcv_nxt. */
@@ -248,25 +248,6 @@ static void dctcp_state(struct sock *sk, u8 new_state)
        }
 }
 
-static void dctcp_update_ack_reserved(struct sock *sk, enum tcp_ca_event ev)
-{
-       struct dctcp *ca = inet_csk_ca(sk);
-
-       switch (ev) {
-       case CA_EVENT_DELAYED_ACK:
-               if (!ca->delayed_ack_reserved)
-                       ca->delayed_ack_reserved = 1;
-               break;
-       case CA_EVENT_NON_DELAYED_ACK:
-               if (ca->delayed_ack_reserved)
-                       ca->delayed_ack_reserved = 0;
-               break;
-       default:
-               /* Don't care for the rest. */
-               break;
-       }
-}
-
 static void dctcp_cwnd_event(struct sock *sk, enum tcp_ca_event ev)
 {
        switch (ev) {
@@ -276,10 +257,6 @@ static void dctcp_cwnd_event(struct sock *sk, enum tcp_ca_event ev)
        case CA_EVENT_ECN_NO_CE:
                dctcp_ce_state_1_to_0(sk);
                break;
-       case CA_EVENT_DELAYED_ACK:
-       case CA_EVENT_NON_DELAYED_ACK:
-               dctcp_update_ack_reserved(sk, ev);
-               break;
        default:
                /* Don't care for the rest. */
                break;
index bea17f1e8302585d70c1e0108ae1c33d149230d8..3b2711e33e4c7c06ed8caec20cf0241f36068f54 100644 (file)
@@ -156,11 +156,24 @@ int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp)
         */
        if (tcptw->tw_ts_recent_stamp &&
            (!twp || (reuse && get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
-               tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2;
-               if (tp->write_seq == 0)
-                       tp->write_seq = 1;
-               tp->rx_opt.ts_recent       = tcptw->tw_ts_recent;
-               tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp;
+               /* In case of repair and re-using TIME-WAIT sockets we still
+                * want to be sure that it is safe as above but honor the
+                * sequence numbers and time stamps set as part of the repair
+                * process.
+                *
+                * Without this check re-using a TIME-WAIT socket with TCP
+                * repair would accumulate a -1 on the repair assigned
+                * sequence number. The first time it is reused the sequence
+                * is -1, the second time -2, etc. This fixes that issue
+                * without appearing to create any others.
+                */
+               if (likely(!tp->repair)) {
+                       tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2;
+                       if (tp->write_seq == 0)
+                               tp->write_seq = 1;
+                       tp->rx_opt.ts_recent       = tcptw->tw_ts_recent;
+                       tp->rx_opt.ts_recent_stamp = tcptw->tw_ts_recent_stamp;
+               }
                sock_hold(sktw);
                return 1;
        }
index 8e08b409c71e1f8e69422f1756d48b5bc55411c3..00e5a300ddb934957463c4abdfbccb50807256e1 100644 (file)
@@ -3523,8 +3523,6 @@ void tcp_send_delayed_ack(struct sock *sk)
        int ato = icsk->icsk_ack.ato;
        unsigned long timeout;
 
-       tcp_ca_event(sk, CA_EVENT_DELAYED_ACK);
-
        if (ato > TCP_DELACK_MIN) {
                const struct tcp_sock *tp = tcp_sk(sk);
                int max_ato = HZ / 2;
@@ -3581,8 +3579,6 @@ void tcp_send_ack(struct sock *sk)
        if (sk->sk_state == TCP_CLOSE)
                return;
 
-       tcp_ca_event(sk, CA_EVENT_NON_DELAYED_ACK);
-
        /* We are not putting this on the write queue, so
         * tcp_transmit_skb() will set the ownership to this
         * sock.
index 0eff75525da101e4fce2798626a317366f94623f..b3885ca22d6fb7aa6165c2773ae02d9885099d8f 100644 (file)
@@ -108,6 +108,7 @@ config IPV6_MIP6
 config IPV6_ILA
        tristate "IPv6: Identifier Locator Addressing (ILA)"
        depends on NETFILTER
+       select DST_CACHE
        select LWTUNNEL
        ---help---
          Support for IPv6 Identifier Locator Addressing (ILA).
index 1323b9679cf718d0023bf5880dcd60fb8602d9db..1c0bb9fb76e61fa7d12317190ebac38847530858 100644 (file)
@@ -799,8 +799,7 @@ static int calipso_opt_update(struct sock *sk, struct ipv6_opt_hdr *hop)
 {
        struct ipv6_txoptions *old = txopt_get(inet6_sk(sk)), *txopts;
 
-       txopts = ipv6_renew_options_kern(sk, old, IPV6_HOPOPTS,
-                                        hop, hop ? ipv6_optlen(hop) : 0);
+       txopts = ipv6_renew_options(sk, old, IPV6_HOPOPTS, hop);
        txopt_put(old);
        if (IS_ERR(txopts))
                return PTR_ERR(txopts);
@@ -1222,8 +1221,7 @@ static int calipso_req_setattr(struct request_sock *req,
        if (IS_ERR(new))
                return PTR_ERR(new);
 
-       txopts = ipv6_renew_options_kern(sk, req_inet->ipv6_opt, IPV6_HOPOPTS,
-                                        new, new ? ipv6_optlen(new) : 0);
+       txopts = ipv6_renew_options(sk, req_inet->ipv6_opt, IPV6_HOPOPTS, new);
 
        kfree(new);
 
@@ -1260,8 +1258,7 @@ static void calipso_req_delattr(struct request_sock *req)
        if (calipso_opt_del(req_inet->ipv6_opt->hopopt, &new))
                return; /* Nothing to do */
 
-       txopts = ipv6_renew_options_kern(sk, req_inet->ipv6_opt, IPV6_HOPOPTS,
-                                        new, new ? ipv6_optlen(new) : 0);
+       txopts = ipv6_renew_options(sk, req_inet->ipv6_opt, IPV6_HOPOPTS, new);
 
        if (!IS_ERR(txopts)) {
                txopts = xchg(&req_inet->ipv6_opt, txopts);
index 5bc2bf3733abd387de8d21932c95ef32eea30d80..20291c2036fcdcd23ccdc2f5b5ae2a1734b2833d 100644 (file)
@@ -1015,29 +1015,21 @@ ipv6_dup_options(struct sock *sk, struct ipv6_txoptions *opt)
 }
 EXPORT_SYMBOL_GPL(ipv6_dup_options);
 
-static int ipv6_renew_option(void *ohdr,
-                            struct ipv6_opt_hdr __user *newopt, int newoptlen,
-                            int inherit,
-                            struct ipv6_opt_hdr **hdr,
-                            char **p)
+static void ipv6_renew_option(int renewtype,
+                             struct ipv6_opt_hdr **dest,
+                             struct ipv6_opt_hdr *old,
+                             struct ipv6_opt_hdr *new,
+                             int newtype, char **p)
 {
-       if (inherit) {
-               if (ohdr) {
-                       memcpy(*p, ohdr, ipv6_optlen((struct ipv6_opt_hdr *)ohdr));
-                       *hdr = (struct ipv6_opt_hdr *)*p;
-                       *p += CMSG_ALIGN(ipv6_optlen(*hdr));
-               }
-       } else {
-               if (newopt) {
-                       if (copy_from_user(*p, newopt, newoptlen))
-                               return -EFAULT;
-                       *hdr = (struct ipv6_opt_hdr *)*p;
-                       if (ipv6_optlen(*hdr) > newoptlen)
-                               return -EINVAL;
-                       *p += CMSG_ALIGN(newoptlen);
-               }
-       }
-       return 0;
+       struct ipv6_opt_hdr *src;
+
+       src = (renewtype == newtype ? new : old);
+       if (!src)
+               return;
+
+       memcpy(*p, src, ipv6_optlen(src));
+       *dest = (struct ipv6_opt_hdr *)*p;
+       *p += CMSG_ALIGN(ipv6_optlen(*dest));
 }
 
 /**
@@ -1063,13 +1055,11 @@ static int ipv6_renew_option(void *ohdr,
  */
 struct ipv6_txoptions *
 ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
-                  int newtype,
-                  struct ipv6_opt_hdr __user *newopt, int newoptlen)
+                  int newtype, struct ipv6_opt_hdr *newopt)
 {
        int tot_len = 0;
        char *p;
        struct ipv6_txoptions *opt2;
-       int err;
 
        if (opt) {
                if (newtype != IPV6_HOPOPTS && opt->hopopt)
@@ -1082,8 +1072,8 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
                        tot_len += CMSG_ALIGN(ipv6_optlen(opt->dst1opt));
        }
 
-       if (newopt && newoptlen)
-               tot_len += CMSG_ALIGN(newoptlen);
+       if (newopt)
+               tot_len += CMSG_ALIGN(ipv6_optlen(newopt));
 
        if (!tot_len)
                return NULL;
@@ -1098,29 +1088,19 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
        opt2->tot_len = tot_len;
        p = (char *)(opt2 + 1);
 
-       err = ipv6_renew_option(opt ? opt->hopopt : NULL, newopt, newoptlen,
-                               newtype != IPV6_HOPOPTS,
-                               &opt2->hopopt, &p);
-       if (err)
-               goto out;
-
-       err = ipv6_renew_option(opt ? opt->dst0opt : NULL, newopt, newoptlen,
-                               newtype != IPV6_RTHDRDSTOPTS,
-                               &opt2->dst0opt, &p);
-       if (err)
-               goto out;
-
-       err = ipv6_renew_option(opt ? opt->srcrt : NULL, newopt, newoptlen,
-                               newtype != IPV6_RTHDR,
-                               (struct ipv6_opt_hdr **)&opt2->srcrt, &p);
-       if (err)
-               goto out;
-
-       err = ipv6_renew_option(opt ? opt->dst1opt : NULL, newopt, newoptlen,
-                               newtype != IPV6_DSTOPTS,
-                               &opt2->dst1opt, &p);
-       if (err)
-               goto out;
+       ipv6_renew_option(IPV6_HOPOPTS, &opt2->hopopt,
+                         (opt ? opt->hopopt : NULL),
+                         newopt, newtype, &p);
+       ipv6_renew_option(IPV6_RTHDRDSTOPTS, &opt2->dst0opt,
+                         (opt ? opt->dst0opt : NULL),
+                         newopt, newtype, &p);
+       ipv6_renew_option(IPV6_RTHDR,
+                         (struct ipv6_opt_hdr **)&opt2->srcrt,
+                         (opt ? (struct ipv6_opt_hdr *)opt->srcrt : NULL),
+                         newopt, newtype, &p);
+       ipv6_renew_option(IPV6_DSTOPTS, &opt2->dst1opt,
+                         (opt ? opt->dst1opt : NULL),
+                         newopt, newtype, &p);
 
        opt2->opt_nflen = (opt2->hopopt ? ipv6_optlen(opt2->hopopt) : 0) +
                          (opt2->dst0opt ? ipv6_optlen(opt2->dst0opt) : 0) +
@@ -1128,37 +1108,6 @@ ipv6_renew_options(struct sock *sk, struct ipv6_txoptions *opt,
        opt2->opt_flen = (opt2->dst1opt ? ipv6_optlen(opt2->dst1opt) : 0);
 
        return opt2;
-out:
-       sock_kfree_s(sk, opt2, opt2->tot_len);
-       return ERR_PTR(err);
-}
-
-/**
- * ipv6_renew_options_kern - replace a specific ext hdr with a new one.
- *
- * @sk: sock from which to allocate memory
- * @opt: original options
- * @newtype: option type to replace in @opt
- * @newopt: new option of type @newtype to replace (kernel-mem)
- * @newoptlen: length of @newopt
- *
- * See ipv6_renew_options().  The difference is that @newopt is
- * kernel memory, rather than user memory.
- */
-struct ipv6_txoptions *
-ipv6_renew_options_kern(struct sock *sk, struct ipv6_txoptions *opt,
-                       int newtype, struct ipv6_opt_hdr *newopt,
-                       int newoptlen)
-{
-       struct ipv6_txoptions *ret_val;
-       const mm_segment_t old_fs = get_fs();
-
-       set_fs(KERNEL_DS);
-       ret_val = ipv6_renew_options(sk, opt, newtype,
-                                    (struct ipv6_opt_hdr __user *)newopt,
-                                    newoptlen);
-       set_fs(old_fs);
-       return ret_val;
 }
 
 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
index 1fb2f3118d60c73433e09f3a71abdbf9a8d92227..d212738e9d100d4e3270f9188466da6b8a3d186c 100644 (file)
@@ -935,20 +935,19 @@ static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
 {
        struct fib6_info *leaf = rcu_dereference_protected(fn->leaf,
                                    lockdep_is_held(&rt->fib6_table->tb6_lock));
-       enum fib_event_type event = FIB_EVENT_ENTRY_ADD;
-       struct fib6_info *iter = NULL, *match = NULL;
+       struct fib6_info *iter = NULL;
        struct fib6_info __rcu **ins;
+       struct fib6_info __rcu **fallback_ins = NULL;
        int replace = (info->nlh &&
                       (info->nlh->nlmsg_flags & NLM_F_REPLACE));
-       int append = (info->nlh &&
-                      (info->nlh->nlmsg_flags & NLM_F_APPEND));
        int add = (!info->nlh ||
                   (info->nlh->nlmsg_flags & NLM_F_CREATE));
        int found = 0;
+       bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
        u16 nlflags = NLM_F_EXCL;
        int err;
 
-       if (append)
+       if (info->nlh && (info->nlh->nlmsg_flags & NLM_F_APPEND))
                nlflags |= NLM_F_APPEND;
 
        ins = &fn->leaf;
@@ -970,8 +969,13 @@ static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
 
                        nlflags &= ~NLM_F_EXCL;
                        if (replace) {
-                               found++;
-                               break;
+                               if (rt_can_ecmp == rt6_qualify_for_ecmp(iter)) {
+                                       found++;
+                                       break;
+                               }
+                               if (rt_can_ecmp)
+                                       fallback_ins = fallback_ins ?: ins;
+                               goto next_iter;
                        }
 
                        if (rt6_duplicate_nexthop(iter, rt)) {
@@ -986,51 +990,71 @@ static int fib6_add_rt2node(struct fib6_node *fn, struct fib6_info *rt,
                                fib6_metric_set(iter, RTAX_MTU, rt->fib6_pmtu);
                                return -EEXIST;
                        }
-
-                       /* first route that matches */
-                       if (!match)
-                               match = iter;
+                       /* If we have the same destination and the same metric,
+                        * but not the same gateway, then the route we try to
+                        * add is sibling to this route, increment our counter
+                        * of siblings, and later we will add our route to the
+                        * list.
+                        * Only static routes (which don't have flag
+                        * RTF_EXPIRES) are used for ECMPv6.
+                        *
+                        * To avoid long list, we only had siblings if the
+                        * route have a gateway.
+                        */
+                       if (rt_can_ecmp &&
+                           rt6_qualify_for_ecmp(iter))
+                               rt->fib6_nsiblings++;
                }
 
                if (iter->fib6_metric > rt->fib6_metric)
                        break;
 
+next_iter:
                ins = &iter->fib6_next;
        }
 
+       if (fallback_ins && !found) {
+               /* No ECMP-able route found, replace first non-ECMP one */
+               ins = fallback_ins;
+               iter = rcu_dereference_protected(*ins,
+                                   lockdep_is_held(&rt->fib6_table->tb6_lock));
+               found++;
+       }
+
        /* Reset round-robin state, if necessary */
        if (ins == &fn->leaf)
                fn->rr_ptr = NULL;
 
        /* Link this route to others same route. */
-       if (append && match) {
+       if (rt->fib6_nsiblings) {
+               unsigned int fib6_nsiblings;
                struct fib6_info *sibling, *temp_sibling;
 
-               if (rt->fib6_flags & RTF_REJECT) {
-                       NL_SET_ERR_MSG(extack,
-                                      "Can not append a REJECT route");
-                      &n