linux-2.6-microblaze.git
2 years agodpaa2-switch: integrate the MAC endpoint support
Ioana Ciornei [Tue, 3 Aug 2021 16:57:43 +0000 (19:57 +0300)]
dpaa2-switch: integrate the MAC endpoint support

Integrate the common MAC endpoint management support into the
dpaa2-switch driver as well. Nothing special happens here, just that the
already available dpaa2-mac functions are also called from dpaa2-switch.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobus: fsl-mc: extend fsl_mc_get_endpoint() to pass interface ID
Ioana Ciornei [Tue, 3 Aug 2021 16:57:42 +0000 (19:57 +0300)]
bus: fsl-mc: extend fsl_mc_get_endpoint() to pass interface ID

In case of a switch DPAA2 object, the interface ID is also needed when
querying for the object endpoint. Extend fsl_mc_get_endpoint() so that
users can also pass the interface ID that are interested in.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodpaa2-switch: no need to check link state right after ndo_open
Ioana Ciornei [Tue, 3 Aug 2021 16:57:41 +0000 (19:57 +0300)]
dpaa2-switch: no need to check link state right after ndo_open

The call to dpaa2_switch_port_link_state_update is a leftover from the
time when on DPAA2 platforms the PHYs were started at boot time so when
an ifconfig was issued on the associated interface, the link status
needed to be checked directly from the ndo_open() callback.  This is not
needed anymore since we are now properly integrated with the PHY layer
thus a link interrupt will come directly from the PHY eventually without
the need to call the sync function.
Fix this up by removing the call to dpaa2_switch_port_link_state_update.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodpaa2-switch: do not enable the DPSW at probe time
Ioana Ciornei [Tue, 3 Aug 2021 16:57:40 +0000 (19:57 +0300)]
dpaa2-switch: do not enable the DPSW at probe time

We should not enable the switch interfaces at probe time since this is
trigged by the open callback. Remove the call dpsw_enable() which does
exactly this.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodpaa2-switch: use the port index in the IRQ handler
Ioana Ciornei [Tue, 3 Aug 2021 16:57:39 +0000 (19:57 +0300)]
dpaa2-switch: use the port index in the IRQ handler

The MC firmware supplies us the switch interface index for which an
interrupt was triggered. Use this to our advantage instead of looping
through all the switch ports and doing unnecessary work.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodpaa2-switch: request all interrupts sources on the DPSW
Ioana Ciornei [Tue, 3 Aug 2021 16:57:38 +0000 (19:57 +0300)]
dpaa2-switch: request all interrupts sources on the DPSW

Request all interrupt sources to be read and then cleared on the DPSW
object. In the next patches we'll also add support for treating other
interrupts.

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: fec: fix MAC internal delay doesn't work
Joakim Zhang [Tue, 3 Aug 2021 05:24:24 +0000 (13:24 +0800)]
net: fec: fix MAC internal delay doesn't work

This patch intends to fix MAC internal delay doesn't work, due to use
of_property_read_u32() incorrectly, and improve this feature a bit:
1) check the delay value if valid, only program register when it's 2000ps.
2) only enable "enet_2x_txclk" clock when require MAC internal delay.

Fixes: fc539459e900 ("net: fec: add MAC internal delayed clock feature support")
Signed-off-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Link: https://lore.kernel.org/r/20210803052424.19008-1-qiangqing.zhang@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: dsa: tag_sja1105: consistently fail with arbitrary input
Vladimir Oltean [Mon, 2 Aug 2021 19:51:37 +0000 (22:51 +0300)]
net: dsa: tag_sja1105: consistently fail with arbitrary input

Dan Carpenter's smatch tests report that the "vid" variable, populated
by sja1105_vlan_rcv when an skb is received by the tagger that has a
VLAN ID which cannot be decoded by tag_8021q, may be uninitialized when
used here:

if (source_port == -1 || switch_id == -1)
skb->dev = dsa_find_designated_bridge_port_by_vid(netdev, vid);

The sja1105 driver, by construction, sets up the switch in a way that
all data plane packets sent towards the CPU port are VLAN-tagged. So it
is practically impossible, in a functional system, for a packet to be
processed by sja1110_rcv() which is not a control packet and does not
have a VLAN header either.

However, it would be nice if the sja1105 tagging driver could
consistently do something valid, for example fail, even if presented with
packets that do not hold valid sja1105 tags. Currently it is a bit hard
to argue that it does that, given the fact that a data plane packet with
no VLAN tag will trigger a call to dsa_find_designated_bridge_port_by_vid
with a vid argument that is an uninitialized stack variable.

To fix this, we can initialize the u16 vid variable with 0, a value that
can never be a bridge VLAN, so dsa_find_designated_bridge_port_by_vid
will always return a NULL skb->dev.

Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://lore.kernel.org/r/20210802195137.303625-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: bridge: switchdev: fix incorrect use of FDB flags when picking the dst device
Vladimir Oltean [Mon, 2 Aug 2021 11:36:33 +0000 (14:36 +0300)]
net: bridge: switchdev: fix incorrect use of FDB flags when picking the dst device

Nikolay points out that it is incorrect to assume that it is impossible
to have an fdb entry with fdb->dst == NULL and the BR_FDB_LOCAL bit in
fdb->flags not set. This is because there are reader-side places that
test_bit(BR_FDB_LOCAL, &fdb->flags) without the br->hash_lock, and if
the updating of the FDB entry happens on another CPU, there are no
memory barriers at writer or reader side which would ensure that the
reader sees the updates to both fdb->flags and fdb->dst in the same
order, i.e. the reader will not see an inconsistent FDB entry.

So we must be prepared to deal with FDB entries where fdb->dst and
fdb->flags are in a potentially inconsistent state, and that means that
fdb->dst == NULL should remain a condition to pick the net_device that
we report to switchdev as being the bridge device, which is what the
code did prior to the blamed patch.

Fixes: 52e4bec15546 ("net: bridge: switchdev: treat local FDBs the same as entries towards the bridge")
Suggested-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Link: https://lore.kernel.org/r/20210802113633.189831-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoRevert "Merge branch 'qcom-dts-updates'"
Alex Elder [Mon, 2 Aug 2021 23:30:19 +0000 (18:30 -0500)]
Revert "Merge branch 'qcom-dts-updates'"

This reverts commit b79c6fba6cd7c49a7dbea9999e182f74cca63e19, reversing
these changes made to 0ac26271344478ff718329fa9d4ef81d4bcbc43b:

  commit 6a0eb6c9d934 ("dt-bindings: net: qcom,ipa: make imem interconnect
                       optional")
  commit f8bd3c82bf7d ("arm64: dts: qcom: sc7280: add IPA information")
  commit fd0f72c34bd9 ("arm64: dts: qcom: sc7180: define ipa_fw_mem node")

I intend for these commits to go through the Qualcomm repository, to
avoid conflicting with other activity being merged there.

Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20210802233019.800250-1-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoocteontx2-af: Fix spelling mistake "Makesure" -> "Make sure"
Colin Ian King [Tue, 3 Aug 2021 10:56:17 +0000 (11:56 +0100)]
octeontx2-af: Fix spelling mistake "Makesure" -> "Make sure"

There is a spelling mistake in a NL_SET_ERR_MSG_MOD message. Fix it.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210803105617.338546-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: decnet: Fix refcount warning for new dn_fib_info
Yajun Deng [Tue, 3 Aug 2021 07:37:39 +0000 (15:37 +0800)]
net: decnet: Fix refcount warning for new dn_fib_info

fib_treeref needs to be set after kzalloc. The old code had a ++ which
led to the confusion when the int was replaced by a refcount_t.

Fixes: 79976892f7ea ("net: convert fib_treeref from int to refcount_t")
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20210803073739.22339-1-yajun.deng@linux.dev
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge branch 'Space-cleanup'
David S. Miller [Tue, 3 Aug 2021 12:05:27 +0000 (13:05 +0100)]
Merge branch 'Space-cleanup'

Arnd Bergmann says:

====================
drivers/net/Space.c cleanup

I discovered that there are still a couple of drivers that rely on
beiong statically initialized from drivers/net/Space.c the way
we did in the last century. As it turns out, there are a couple
of simplifications that can be made here, as well as some minor
bugfixes.

There are four classes of drivers that use this:

- most 10mbit ISA bus ethernet drivers (and one 100mbit one)
- both ISA localtalk drivers
- several m68k ethernet drivers
- one obsolete WAN driver

I found that the drivers using in arch/m68k/ don't actually benefit
from being probed this way as they do not rely on the netdev= command
line arguments, they have simply never been changed to work like a
modern driver.

I had previously sent a patch to remove the sbni/granch driver, and
there were no objections to this patch but forgot to resend it after
some discussion about another patch in the same series.

For the ISA drivers, there is usually no way to probe multiple devices
at boot time other than the netdev= arguments, so all that logic is left
in place for the moment, but centralized in a single file that only gets
included in the kernel build if one or more of the drivers are built-in.

I'm also changing the old-style init_module() functions in these drivers
to static functions with a module_init() annotation, to more closely
resemble modern drivers. These are the last drivers in the kernel to
still use init_module/cleanup_module, removing those may enable future
cleanups to the module loading process.

       Arnd

Changes in v2:

- replace xsurf100 change with Michael's version
- make it PATCH instead of RFC
- rebase to net-next as of August 3
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoethernet: isa: convert to module_init/module_exit
Arnd Bergmann [Tue, 3 Aug 2021 11:40:51 +0000 (13:40 +0200)]
ethernet: isa: convert to module_init/module_exit

There are a couple of ISA ethernet drivers that use the old
init_module/cleanup_module function names for the main entry
points, nothing else uses those any more.

Change them to the documented method with module_init()
and module_exit() markers next to static functions.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agowan: hostess_sv11: use module_init/module_exit helpers
Arnd Bergmann [Tue, 3 Aug 2021 11:40:50 +0000 (13:40 +0200)]
wan: hostess_sv11: use module_init/module_exit helpers

This is one of very few drivers using the old init_module/cleanup_module
function names. Change it over to the modern method.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agowan: remove sbni/granch driver
Arnd Bergmann [Tue, 3 Aug 2021 11:40:49 +0000 (13:40 +0200)]
wan: remove sbni/granch driver

The driver was merged in 1999 and has only ever seen treewide cleanups
since then, with no indication whatsoever that anyone has actually
had access to hardware for testing the patches.

>From the information in the link below, it appears that the hardware
is for some leased line system in Russia that has since been
discontinued, and useless without any remote end to connect to.

As the driver still feels like a Linux-2.2 era artifact today, it
appears that the best way forward is to just delete it.

Link: https://www.tms.ru/%D0%90%D0%B4%D0%B0%D0%BF%D1%82%D0%B5%D1%80_%D0%B4%D0%BB%D1%8F_%D0%B2%D1%8B%D0%B4%D0%B5%D0%BB%D0%B5%D0%BD%D0%BD%D1%8B%D1%85_%D0%BB%D0%B8%D0%BD%D0%B8%D0%B9_Granch_SBNI12-10
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agowan: remove stale Kconfig entries
Arnd Bergmann [Tue, 3 Aug 2021 11:40:48 +0000 (13:40 +0200)]
wan: remove stale Kconfig entries

The dscc4 driver was removed in 2019 but these Kconfig entries remain,
so remove them as well.

Fixes: 28c9eb9042a9 ("net/wan: dscc4: remove broken dscc4 driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomake legacy ISA probe optional
Arnd Bergmann [Tue, 3 Aug 2021 11:40:47 +0000 (13:40 +0200)]
make legacy ISA probe optional

There are very few ISA drivers left that rely on the static probing from
drivers/net/Space.o. Make them all select a new CONFIG_NETDEV_LEGACY_INIT
symbol, and drop the entire probe logic when that is disabled.

The 9 drivers that are called from Space.c are the same set that
calls netdev_boot_setup_check().

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agomove netdev_boot_setup into Space.c
Arnd Bergmann [Tue, 3 Aug 2021 11:40:46 +0000 (13:40 +0200)]
move netdev_boot_setup into Space.c

This is now only used by a handful of old ISA drivers,
and can be moved into the file they already all depend on.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoxsurf100: drop include of lib8390.c
Michael Schmitz [Tue, 3 Aug 2021 11:40:45 +0000 (13:40 +0200)]
xsurf100: drop include of lib8390.c

Now that ax88796.c exports the ax_NS8390_reinit() symbol, we can
include 8390.h instead of lib8390.c, avoiding duplication of that
function and killing a few compile warnings in the bargain.

Fixes: 861928f4e60e826c ("net-next: New ax88796 platform
driver for Amiga X-Surf 100 Zorro board (m68k)")

Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoax88796: export ax_NS8390_init() hook
Michael Schmitz [Tue, 3 Aug 2021 11:40:44 +0000 (13:40 +0200)]
ax88796: export ax_NS8390_init() hook

The block I/O code for the new X-Surf 100 ax88796 driver needs
ax_NS8390_init() for error fixup in its block_output function.

Export this static function through the ax_NS8390_reinit()
wrapper so we can lose the lib8380.c include in the X-Surf 100
driver.

[arnd: add the declaration in the header to avoid a
 -Wmissing-prototypes warning]
Fixes: 861928f4e60e826c ("net-next: New ax88796 platform
driver for Amiga X-Surf 100 Zorro board (m68k)")
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agom68k: remove legacy probing
Arnd Bergmann [Tue, 3 Aug 2021 11:40:43 +0000 (13:40 +0200)]
m68k: remove legacy probing

There are six m68k specific drivers that use the legacy probe method
in drivers/net/Space.c. However, all of these only support a single
device, and they completely ignore the command line settings from
netdev_boot_setup_check, so there is really no point at all.

Aside from sun3_82586, these already have a module_init function that
can be used for built-in mode as well, simply by removing the #ifdef.

Note that the 82596 driver was previously used on ISA as well, but
that got dropped long ago.

Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agocs89x0: rework driver configuration
Arnd Bergmann [Tue, 3 Aug 2021 11:40:42 +0000 (13:40 +0200)]
cs89x0: rework driver configuration

There are two drivers in the cs89x0 file, with the CONFIG_CS89x0_PLATFORM
symbol deciding which one is getting built. This is somewhat confusing
and makes it more likely ton configure a driver that works nowhere.

Split up the Kconfig option into separate ISA and PLATFORM drivers,
with the ISA symbol explicitly connecting to the static probing in
drivers/net/Space.c

The two drivers are still mutually incompatible at compile time,
which could be lifted by splitting them into multiple files,
but in practice this will make no difference.

The platform driver can now be enabled for compile-testing on
non-ARM machines.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years ago3c509: stop calling netdev_boot_setup_check
Arnd Bergmann [Tue, 3 Aug 2021 11:40:41 +0000 (13:40 +0200)]
3c509: stop calling netdev_boot_setup_check

This driver never uses the information returned by
netdev_boot_setup_check, and is not called by the boot-time probing from
driver/net/Space.c, so just remove these stale references.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoappletalk: ltpc: remove static probing
Arnd Bergmann [Tue, 3 Aug 2021 11:40:40 +0000 (13:40 +0200)]
appletalk: ltpc: remove static probing

This driver never relies on the netdev_boot_setup_check()
to get its configuration, so it can just as well do its
own probing all the time.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonatsemi: sonic: stop calling netdev_boot_setup_check
Arnd Bergmann [Tue, 3 Aug 2021 11:40:39 +0000 (13:40 +0200)]
natsemi: sonic: stop calling netdev_boot_setup_check

The data from the kernel command line is no longer used since the
probe function gets it from the platform device resources instead.

The jazz version was changed to be like this in 2007, the xtensa
version apparently copied the code from there.

Fixes: ed9f0e0bf3ce ("remove setup of platform device from jazzsonic.c")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobcmgenet: remove call to netdev_boot_setup_check
Arnd Bergmann [Tue, 3 Aug 2021 11:40:38 +0000 (13:40 +0200)]
bcmgenet: remove call to netdev_boot_setup_check

The driver has never used the netdev->{irq,base_addr,mem_start}
members, so this call is completely unnecessary.

Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'ethtool-runtime-pm'
David S. Miller [Tue, 3 Aug 2021 11:58:22 +0000 (12:58 +0100)]
Merge branch 'ethtool-runtime-pm'

Heiner Kallweit says:

====================
ethtool: runtime-resume netdev parent before ethtool ops

If a network device is runtime-suspended then:
- network device may be flagged as detached and all ethtool ops (even if
  not accessing the device) will fail because netif_device_present()
  returns false
- ethtool ops may fail because device is not accessible (e.g. because being
  in D3 in case of a PCI device)

It may not be desirable that userspace can't use even simple ethtool ops
that not access the device if interface or link is down. To be more friendly
to userspace let's ensure that device is runtime-resumed when executing
ethtool ops in kernel.

This patch series covers the typical case that the netdev parent is power-
managed, e.g. a PCI device. Not sure whether cases exist where the netdev
itself is power-managed. If yes then we may need an extension for this.
But the series as-is at least shouldn't cause problems in that case.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoethtool: runtime-resume netdev parent in ethnl_ops_begin
Heiner Kallweit [Sun, 1 Aug 2021 10:41:31 +0000 (12:41 +0200)]
ethtool: runtime-resume netdev parent in ethnl_ops_begin

If a network device is runtime-suspended then:
- network device may be flagged as detached and all ethtool ops (even if not
  accessing the device) will fail because netif_device_present() returns
  false
- ethtool ops may fail because device is not accessible (e.g. because being
  in D3 in case of a PCI device)

It may not be desirable that userspace can't use even simple ethtool ops
that not access the device if interface or link is down. To be more friendly
to userspace let's ensure that device is runtime-resumed when executing the
respective ethtool op in kernel.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoethtool: move netif_device_present check from ethnl_parse_header_dev_get to ethnl_ops...
Heiner Kallweit [Sun, 1 Aug 2021 10:40:05 +0000 (12:40 +0200)]
ethtool: move netif_device_present check from ethnl_parse_header_dev_get to ethnl_ops_begin

If device is runtime-suspended and not accessible then it may be
flagged as not present. If checking whether device is present is
done too early then we may bail out before we have the chance to
runtime-resume the device. Therefore move this check to
ethnl_ops_begin(). This is in preparation of a follow-up patch
that tries to runtime-resume the device before executing ethtool
ops.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoethtool: move implementation of ethnl_ops_begin/complete to netlink.c
Heiner Kallweit [Sun, 1 Aug 2021 10:37:39 +0000 (12:37 +0200)]
ethtool: move implementation of ethnl_ops_begin/complete to netlink.c

In preparation of subsequent extensions to both functions move the
implementations from netlink.h to netlink.c.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoethtool: runtime-resume netdev parent before ethtool ioctl ops
Heiner Kallweit [Sun, 1 Aug 2021 10:36:48 +0000 (12:36 +0200)]
ethtool: runtime-resume netdev parent before ethtool ioctl ops

If a network device is runtime-suspended then:
- network device may be flagged as detached and all ethtool ops (even if not
  accessing the device) will fail because netif_device_present() returns
  false
- ethtool ops may fail because device is not accessible (e.g. because being
  in D3 in case of a PCI device)

It may not be desirable that userspace can't use even simple ethtool ops
that not access the device if interface or link is down. To be more friendly
to userspace let's ensure that device is runtime-resumed when executing the
respective ethtool op in kernel.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agovirtio-net: realign page_to_skb() after merges
Jakub Kicinski [Mon, 2 Aug 2021 17:57:29 +0000 (10:57 -0700)]
virtio-net: realign page_to_skb() after merges

We ended up merging two versions of the same patch set:

commit 8fb7da9e9907 ("virtio_net: get build_skb() buf by data ptr")
commit 5c37711d9f27 ("virtio-net: fix for unable to handle page fault for address")

into net, and

commit 7bf64460e3b2 ("virtio-net: get build_skb() buf by data ptr")
commit 6c66c147b9a4 ("virtio-net: fix for unable to handle page fault for address")

into net-next. Redo the merge from commit 126285651b7f ("Merge
ra.kernel.org:/pub/scm/linux/kernel/git/netdev/net"), so that
the most recent code remains.

Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'bnxt_en-rx-ring'
David S. Miller [Tue, 3 Aug 2021 11:38:47 +0000 (12:38 +0100)]
Merge branch 'bnxt_en-rx-ring'

Michael Chan says:

====================
bnxt_en: Increase maximum RX ring size when jumbo ring is unused

The RX jumbo ring is automatically enabled when HW GRO/LRO is enabled or
when the MTU exceeds the page size.  The RX jumbo ring provides a lot
more RX buffer space when it is in use.  When the RX jumbo ring is not
in use, some users report that the current maximum of 2K buffers is
too limiting.  This patchset increases the maximum to 8K buffers when
the RX jumbo ring is not used.  The default RX ring size is unchanged
at 511.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobnxt_en: Increase maximum RX ring size if jumbo ring is not used
Michael Chan [Mon, 2 Aug 2021 14:52:39 +0000 (10:52 -0400)]
bnxt_en: Increase maximum RX ring size if jumbo ring is not used

The current maximum RX ring size is defined assuming the RX jumbo ring
(aka aggregation ring) is used.  The RX jumbo ring is automicatically used
when the MTU exceeds a threshold or when rx-gro-hw/lro is enabled.  The RX
jumbo ring is automatically sized up to 4 times the size of the RX ring
size.

The BNXT_MAX_RX_DESC_CNT constant is the upper limit on the size of the
RX ring whether or not the RX jumbo ring is used.  Obviously, the
maximum amount of RX buffer space is significantly less when the RX jumbo
ring is not used.

To increase flexibility for the user who does not use the RX jumbo ring,
we now define a bigger maximum RX ring size when the RX jumbo ring is not
used.  The maximum RX ring size is now up to 8K when the RX jumbo ring
is not used.  The maximum completion ring size also needs to be scaled
up to accomodate the larger maximum RX ring size.

Note that when the RX jumbo ring is re-enabled, the RX ring size will
automatically drop if it exceeds the maximum.

Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobnxt_en: Don't use static arrays for completion ring pages
Michael Chan [Mon, 2 Aug 2021 14:52:38 +0000 (10:52 -0400)]
bnxt_en: Don't use static arrays for completion ring pages

We currently store these page addresses and DMA addreses in static
arrays.  On systems with 4K pages, we support up to 64 pages per
completion ring.  The actual number of pages for each completion ring
may be much less than 64.  For example, when the RX ring size is set
to the default 511 entries, only 16 completion ring pages are needed
per ring.

In the next patch, we'll be doubling the maximum number of completion
pages.  So we convert to allocate these arrays as needed instead of
declaring them statically.

Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Edwin Peer <edwin.peer@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: Keep vertical alignment
Yajun Deng [Mon, 2 Aug 2021 08:05:08 +0000 (16:05 +0800)]
net: Keep vertical alignment

Those files under /proc/net/stat/ don't have vertical alignment, it looks
very difficult. Modify the seq_printf statement, keep vertical alignment.

v2:
 - Use seq_puts() and seq_printf() correctly.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Yajun Deng <yajun.deng@linux.dev>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobonding: add new option lacp_active
Hangbin Liu [Mon, 2 Aug 2021 03:02:19 +0000 (11:02 +0800)]
bonding: add new option lacp_active

Add an option lacp_active, which is similar with team's runner.active.
This option specifies whether to send LACPDU frames periodically. If set
on, the LACPDU frames are sent along with the configured lacp_rate
setting. If set off, the LACPDU frames acts as "speak when spoken to".

Note, the LACPDU state frames still will be sent when init or unbind port.

v2: remove module parameter

Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoqed: Remove duplicated include of kernel.h
zhouchuangao [Mon, 2 Aug 2021 02:18:38 +0000 (19:18 -0700)]
qed: Remove duplicated include of kernel.h

Duplicate include header file <linux/kernel.h>
line 4: #include <linux/kernel.h>
line 7: #include <linux/kernel.h>

Signed-off-by: zhouchuangao <zhouchuangao@vivo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agodrivers/net/usb: Remove all strcpy() uses
Len Baker [Sun, 1 Aug 2021 17:12:26 +0000 (19:12 +0200)]
drivers/net/usb: Remove all strcpy() uses

strcpy() performs no bounds checking on the destination buffer. This
could result in linear overflows beyond the end of the buffer, leading
to all kinds of misbehaviors. The safe replacement is strscpy().

Signed-off-by: Len Baker <len.baker@gmx.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoqed: Remove redundant prints from the iWARP SYN handling
Shai Malin [Sun, 1 Aug 2021 10:28:40 +0000 (13:28 +0300)]
qed: Remove redundant prints from the iWARP SYN handling

Remove redundant prints from the iWARP SYN handling.

Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Shai Malin <smalin@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoqed: Skip DORQ attention handling during recovery
Shai Malin [Sun, 1 Aug 2021 10:26:38 +0000 (13:26 +0300)]
qed: Skip DORQ attention handling during recovery

The device recovery flow will reset the entire HW device, in that case
the DORQ HW block attention is redundant.

Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Shai Malin <smalin@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoqed: Avoid db_recovery during recovery
Shai Malin [Sun, 1 Aug 2021 10:23:40 +0000 (13:23 +0300)]
qed: Avoid db_recovery during recovery

Avoid calling the qed doorbell recovery - qed_db_rec_handler()
during device recovery.

Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Shai Malin <smalin@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'skb_expand_head'
David S. Miller [Tue, 3 Aug 2021 10:21:39 +0000 (11:21 +0100)]
Merge branch 'skb_expand_head'

Vasily Averin says:

====================
skbuff: introduce skb_expand_head()

currently if skb does not have enough headroom skb_realloc_headrom is called.
It is not optimal because it creates new skb.

this patch set introduces new helper skb_expand_head()
Unlike skb_realloc_headroom, it does not allocate a new skb if possible;
copies skb->sk on new skb when as needed and frees original skb in case of failures.

This helps to simplify ip[6]_finish_output2(), ip6_xmit() and few other
functions in vrf, ax25 and bpf.

There are few other cases where this helper can be used
but it requires an additional investigations.

v3 changes:
 - ax25 compilation warning fixed
 - v5.14-rc4 rebase
 - now it does not depend on non-committed pathces

v2 changes:
 - helper's name was changed to skb_expand_head
 - fixed few mistakes inside skb_expand_head():
    skb_set_owner_w should set sk on nskb
    kfree was replaced by kfree_skb()
    improved warning message
 - added minor refactoring in changed functions in vrf and bpf patches
 - removed kfree_skb() in ax25_rt_build_path caller ax25_ip_xmit
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobpf: use skb_expand_head in bpf_out_neigh_v4/6
Vasily Averin [Mon, 2 Aug 2021 08:52:54 +0000 (11:52 +0300)]
bpf: use skb_expand_head in bpf_out_neigh_v4/6

Unlike skb_realloc_headroom, new helper skb_expand_head
does not allocate a new skb if possible.

Additionally this patch replaces commonly used dereferencing with variables.

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoax25: use skb_expand_head
Vasily Averin [Mon, 2 Aug 2021 08:52:47 +0000 (11:52 +0300)]
ax25: use skb_expand_head

Use skb_expand_head() in ax25_transmit_buffer and ax25_rt_build_path.
Unlike skb_realloc_headroom, new helper does not allocate a new skb if possible.

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agovrf: use skb_expand_head in vrf_finish_output
Vasily Averin [Mon, 2 Aug 2021 08:52:40 +0000 (11:52 +0300)]
vrf: use skb_expand_head in vrf_finish_output

Unlike skb_realloc_headroom, new helper skb_expand_head
does not allocate a new skb if possible.

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv4: use skb_expand_head in ip_finish_output2
Vasily Averin [Mon, 2 Aug 2021 08:52:35 +0000 (11:52 +0300)]
ipv4: use skb_expand_head in ip_finish_output2

Unlike skb_realloc_headroom, new helper skb_expand_head
does not allocate a new skb if possible.

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv6: use skb_expand_head in ip6_xmit
Vasily Averin [Mon, 2 Aug 2021 08:52:29 +0000 (11:52 +0300)]
ipv6: use skb_expand_head in ip6_xmit

Unlike skb_realloc_headroom, new helper skb_expand_head
does not allocate a new skb if possible.

Additionally this patch replaces commonly used dereferencing with variables.

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoipv6: use skb_expand_head in ip6_finish_output2
Vasily Averin [Mon, 2 Aug 2021 08:52:22 +0000 (11:52 +0300)]
ipv6: use skb_expand_head in ip6_finish_output2

Unlike skb_realloc_headroom, new helper skb_expand_head does not allocate
a new skb if possible.

Additionally this patch replaces commonly used dereferencing with variables.

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoskbuff: introduce skb_expand_head()
Vasily Averin [Mon, 2 Aug 2021 08:52:15 +0000 (11:52 +0300)]
skbuff: introduce skb_expand_head()

Like skb_realloc_headroom(), new helper increases headroom of specified skb.
Unlike skb_realloc_headroom(), it does not allocate a new skb if possible;
copies skb->sk on new skb when as needed and frees original skb in case
of failures.

This helps to simplify ip[6]_finish_output2() and a few other similar cases.

Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge tag 'mlx5-updates-2021-08-02' of git://git.kernel.org/pub/scm/linux/kernel...
David S. Miller [Tue, 3 Aug 2021 10:16:13 +0000 (11:16 +0100)]
Merge tag 'mlx5-updates-2021-08-02' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
This patch-set changes the TTC (Traffic Type Classification) logic
to be independent from the mlx5 ethernet driver by renaming the traffic
types enums and making the TTC API generic to the mlx5 core driver.

It allows to decouple TTC logic from mlx5e and reused by other parts
of mlx5 drivers, namely ADQ and lag TX steering hashing.

Patches overview:
1 - Rename traffic type enums to be mlx5 generic.
2 - Rename related TTC arguments and functions.
3 - Remove dependency in the mlx5e driver from the TTC implementation.
4 - Move TTC logic to fs_ttc.
5 - Embed struct mlx5_ttc_table in fs_ttc.

The refactoring series is followed by misc' cleanup patches.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet/mlx5: Fix missing return value in mlx5_devlink_eswitch_inline_mode_set()
Jiapeng Chong [Thu, 22 Jul 2021 09:58:16 +0000 (17:58 +0800)]
net/mlx5: Fix missing return value in mlx5_devlink_eswitch_inline_mode_set()

The return value is missing in this code scenario, add the return value
'0' to the return value 'err'.

Eliminate the follow smatch warning:

drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c:3083
mlx5_devlink_eswitch_inline_mode_set() warn: missing error code 'err'.

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Fixes: 8e0aa4bc959c ("net/mlx5: E-switch, Protect eswitch mode changes")
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Return -EOPNOTSUPP if more relevant when parsing tc actions
Roi Dayan [Mon, 26 Jul 2021 12:17:31 +0000 (15:17 +0300)]
net/mlx5e: Return -EOPNOTSUPP if more relevant when parsing tc actions

Instead of returning -EINVAL.

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Remove redundant assignment of counter to null
Roi Dayan [Mon, 26 Jul 2021 12:13:35 +0000 (15:13 +0300)]
net/mlx5e: Remove redundant assignment of counter to null

counter is being initialized before being used.

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Remove redundant parse_attr arg
Roi Dayan [Mon, 26 Jul 2021 12:11:43 +0000 (15:11 +0300)]
net/mlx5e: Remove redundant parse_attr arg

Passing parse_attr is redundant in parse_tc_nic_actions() and
mlx5e_tc_add_nic_flow() as we can get it from flow.
This is the same as with parse_tc_fdb_actions() and mlx5e_tc_add_fdb_flow().

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Remove redundant cap check for flow counter
Roi Dayan [Sun, 25 Jul 2021 13:04:00 +0000 (16:04 +0300)]
net/mlx5e: Remove redundant cap check for flow counter

The cap is very old and today will always exists.
The cap is not being checked anywhere else. Remove the check from
drop action when parsing tc rules in nic mode.

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Remove redundant filter_dev arg from parse_tc_fdb_actions()
Roi Dayan [Thu, 22 Jul 2021 09:10:54 +0000 (12:10 +0300)]
net/mlx5e: Remove redundant filter_dev arg from parse_tc_fdb_actions()

filter_dev is saved in parse_attr. and being used in other cases from
there. use it also for the leftover case.

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Remove redundant tc act includes
Roi Dayan [Thu, 22 Jul 2021 08:58:08 +0000 (11:58 +0300)]
net/mlx5e: Remove redundant tc act includes

Since the code changed to use the flow action infra
there is no usage of tcf values from those includes.
Remove those.

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5: Embed mlx5_ttc_table
Maor Gottlieb [Sun, 18 Jul 2021 12:53:53 +0000 (15:53 +0300)]
net/mlx5: Embed mlx5_ttc_table

mlx5_ttc_table struct shouldn't be exposed to the users so
this patch make it internal to ttc.

In addition add a getter function to get the TTC flow table for users
that need to add a rule which points on it.

Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
2 years agonet/mlx5: Move TTC logic to fs_ttc
Maor Gottlieb [Fri, 2 Jul 2021 07:38:32 +0000 (10:38 +0300)]
net/mlx5: Move TTC logic to fs_ttc

Now that TTC logic is not dependent on mlx5e structs, move it to
lib/fs_ttc.c so it could be used other part of the mlx5 driver.

Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Decouple TTC logic from mlx5e
Maor Gottlieb [Fri, 2 Jul 2021 11:25:14 +0000 (14:25 +0300)]
net/mlx5e: Decouple TTC logic from mlx5e

Remove dependency in the mlx5e driver from the TTC implementation
by changing the TTC related functions to receive mlx5 generic arguments.
It allows to decouple TTC logic from mlx5e and reused by other parts of
mlx5 driver.

Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Rename some related TTC args and functions
Maor Gottlieb [Fri, 2 Jul 2021 11:49:37 +0000 (14:49 +0300)]
net/mlx5e: Rename some related TTC args and functions

Since TTC logic is going to be moved to a separate file, make the
relevant functions and arguments that used by TTC to be mlx5 generic.

Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Rename traffic type enums
Maor Gottlieb [Fri, 2 Jul 2021 08:42:28 +0000 (11:42 +0300)]
net/mlx5e: Rename traffic type enums

Rename traffic type enums as part of the preparation for moving
the traffic type logic to a separate file.

Signed-off-by: Maor Gottlieb <maorg@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Allocate the array of channels according to the real max_nch
Maxim Mikityanskiy [Mon, 12 Apr 2021 16:10:17 +0000 (19:10 +0300)]
net/mlx5e: Allocate the array of channels according to the real max_nch

The channels array in struct mlx5e_rx_res is converted to a dynamic one,
which will use the dynamic value of max_nch instead of
implementation-defined maximum of MLX5E_MAX_NUM_CHANNELS.

Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Hide all implementation details of mlx5e_rx_res
Maxim Mikityanskiy [Fri, 9 Apr 2021 16:01:51 +0000 (19:01 +0300)]
net/mlx5e: Hide all implementation details of mlx5e_rx_res

This commit moves all implementation details of struct mlx5e_rx_res
under en/rx_res.c. All access to RX resources is now done using methods.
Encapsulating RX resources into an object allows for better
manageability, because all the implementation details are now in a
single place, and external code can use only a limited set of API
methods to init/teardown the whole thing, reconfigure RSS and LRO
parameters, connect TIRs to flow steering and activate/deactivate TIRs.

mlx5e_rx_res is self-contained and doesn't depend on struct mlx5e_priv
or include en.h.

Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Introduce mlx5e_channels API to get RQNs
Maxim Mikityanskiy [Fri, 9 Apr 2021 14:31:09 +0000 (17:31 +0300)]
net/mlx5e: Introduce mlx5e_channels API to get RQNs

Currently, struct mlx5e_channels is defined in en.h, along with a lot of
other stuff. In the following commit mlx5e_rx_res will need to get RQNs
(RQ hardware IDs), given a pointer to mlx5e_channels and the channel
index. In order to make it possible without including the whole en.h,
this commit introduces functions that will hide the implementation
details of mlx5e_channels.

Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: Use a new initializer to build uniform indir table
Maxim Mikityanskiy [Thu, 8 Apr 2021 09:16:34 +0000 (12:16 +0300)]
net/mlx5e: Use a new initializer to build uniform indir table

Replace mlx5e_build_default_indir_rqt with a new initializer of struct
mlx5e_rss_params_indir that works directly with the struct, rather than
its internals.

The new initializer is called mlx5e_rss_params_indir_init_uniform, which
also reflects the purpose (uniform spreading) better.

Signed-off-by: Maxim Mikityanskiy <maximmi@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx4: make the array states static const, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 15:37:42 +0000 (16:37 +0100)]
net/mlx4: make the array states static const, makes object smaller

Don't populate the array states on the stack but instead it
static const. Makes the object code smaller by 79 bytes.

Before:
   text   data   bss    dec    hex filename
  21309   8304   192  29805   746d drivers/net/ethernet/mellanox/mlx4/qp.o

After:
   text   data   bss    dec    hex filename
  21166   8368   192  29726   741e drivers/net/ethernet/mellanox/mlx4/qp.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://lore.kernel.org/r/20210801153742.147304-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: 3c509: make the array if_names static const, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 15:26:50 +0000 (16:26 +0100)]
net: 3c509: make the array if_names static const, makes object smaller

Don't populate the array if_names on the stack but instead it
static const. Makes the object code smaller by 99 bytes.

Before:
   text    data     bss     dec     hex filename
  27886   10752     672   39310    998e ./drivers/net/ethernet/3com/3c509.o

After:
   text    data     bss     dec     hex filename
  27723   10816     672   39211    992b ./drivers/net/ethernet/3com/3c509.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210801152650.146572-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agodpaa2-eth: make the array faf_bits static const, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 15:22:09 +0000 (16:22 +0100)]
dpaa2-eth: make the array faf_bits static const, makes object smaller

Don't populate the array faf_bits on the stack but instead it
static const. Makes the object code smaller by 175 bytes.

Before:
   text  data   bss     dec   hex filename
   9645  4552     0   14197  3775 ../freescale/dpaa2/dpaa2-eth-devlink.o

After:
   text  data   bss     dec   hex filename
   9406  4616     0   14022  36c6 ../freescale/dpaa2/dpaa2-eth-devlink.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210801152209.146359-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoqlcnic: make the array random_data static const, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 15:16:59 +0000 (16:16 +0100)]
qlcnic: make the array random_data static const, makes object smaller

Don't populate the array random_data on the stack but instead it
static const. Makes the object code smaller by 66 bytes.

Before:
   text    data     bss     dec     hex filename
  52895   10976       0   63871    f97f ../qlogic/qlcnic/qlcnic_ethtool.o

After:
   text    data     bss     dec     hex filename
  52701   11104       0   63805    f93d ../qlogic//qlcnic/qlcnic_ethtool.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210801151659.146113-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: marvell: make the array name static, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 15:06:47 +0000 (16:06 +0100)]
net: marvell: make the array name static, makes object smaller

Don't populate the const array name on the stack but instead it
static. Makes the object code smaller by 28 bytes. Add a missing
const to clean up a checkpatch warning.

Before:
   text    data   bss     dec     hex filename
 124565   31565   384  156514   26362 drivers/net/ethernet/marvell/sky2.o

After:
   text    data   bss     dec     hex filename
 124441   31661   384  156486   26346 drivers/net/ethernet/marvell/sky2.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210801150647.145728-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agocxgb4: make the array match_all_mac static, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 15:12:05 +0000 (16:12 +0100)]
cxgb4: make the array match_all_mac static, makes object smaller

Don't populate the array match_all_mac on the stack but instead it
static const. Makes the object code smaller by 75 bytes.

Before:
   text    data     bss     dec     hex filename
  46701    8960      64   55725    d9ad ../chelsio/cxgb4/cxgb4_filter.o

After:
   text    data     bss     dec     hex filename
  46338    9120     192   55650    d962 ../chelsio/cxgb4/cxgb4_filter.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210801151205.145924-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoipv4: Fix refcount warning for new fib_info
David Ahern [Mon, 2 Aug 2021 16:02:21 +0000 (10:02 -0600)]
ipv4: Fix refcount warning for new fib_info

Ioana reported a refcount warning when booting over NFS:

[    5.042532] ------------[ cut here ]------------
[    5.047184] refcount_t: addition on 0; use-after-free.
[    5.052324] WARNING: CPU: 7 PID: 1 at lib/refcount.c:25 refcount_warn_saturate+0xa4/0x150
...
[    5.167201] Call trace:
[    5.169635]  refcount_warn_saturate+0xa4/0x150
[    5.174067]  fib_create_info+0xc00/0xc90
[    5.177982]  fib_table_insert+0x8c/0x620
[    5.181893]  fib_magic.isra.0+0x110/0x11c
[    5.185891]  fib_add_ifaddr+0xb8/0x190
[    5.189629]  fib_inetaddr_event+0x8c/0x140

fib_treeref needs to be set after kzalloc. The old code had a ++ which
led to the confusion when the int was replaced by a refcount_t.

Fixes: 79976892f7ea ("net: convert fib_treeref from int to refcount_t")
Signed-off-by: David Ahern <dsahern@kernel.org>
Reported-by: Ioana Ciornei <ciorneiioana@gmail.com>
Cc: Yajun Deng <yajun.deng@linux.dev>
Tested-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Link: https://lore.kernel.org/r/20210802160221.27263-1-dsahern@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: phy: mscc: make some arrays static const, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 07:01:55 +0000 (08:01 +0100)]
net: phy: mscc: make some arrays static const, makes object smaller

Don't populate arrays on the stack but instead them static const.
Makes the object code smaller by 280 bytes.

Before:
   text    data     bss     dec     hex filename
  24142    4368     192   28702    701e ./drivers/net/phy/mscc/mscc_ptp.o

After:
   text    data     bss     dec     hex filename
  23830    4400     192   28422    6f06 ./drivers/net/phy/mscc/mscc_ptp.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210801070155.139057-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonetdevsim: make array res_ids static const, makes object smaller
Colin Ian King [Sun, 1 Aug 2021 06:53:28 +0000 (07:53 +0100)]
netdevsim: make array res_ids static const, makes object smaller

Don't populate the array res_ids on the stack but instead it
static const. Makes the object code smaller by 14 bytes.

Before:
   text    data     bss     dec     hex filename
  50833    8314     256   59403    e80b ./drivers/net/netdevsim/fib.o

After:
   text    data     bss     dec     hex filename
  50755    8378     256   59389    e7fd ./drivers/net/netdevsim/fib.o

(gcc version 10.2.0)

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210801065328.138906-1-colin.king@canonical.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/ipv4: Replace one-element array with flexible-array member
Gustavo A. R. Silva [Sat, 31 Jul 2021 17:08:30 +0000 (12:08 -0500)]
net/ipv4: Replace one-element array with flexible-array member

There is a regular need in the kernel to provide a way to declare having
a dynamically sized set of trailing elements in a structure. Kernel code
should always use “flexible array members”[1] for these cases. The older
style of one-element or zero-length arrays should no longer be used[2].

Use an anonymous union with a couple of anonymous structs in order to
keep userspace unchanged:

$ pahole -C ip_msfilter net/ipv4/ip_sockglue.o
struct ip_msfilter {
union {
struct {
__be32     imsf_multiaddr_aux;   /*     0     4 */
__be32     imsf_interface_aux;   /*     4     4 */
__u32      imsf_fmode_aux;       /*     8     4 */
__u32      imsf_numsrc_aux;      /*    12     4 */
__be32     imsf_slist[1];        /*    16     4 */
};                                       /*     0    20 */
struct {
__be32     imsf_multiaddr;       /*     0     4 */
__be32     imsf_interface;       /*     4     4 */
__u32      imsf_fmode;           /*     8     4 */
__u32      imsf_numsrc;          /*    12     4 */
__be32     imsf_slist_flex[0];   /*    16     0 */
};                                       /*     0    16 */
};                                               /*     0    20 */

/* size: 20, cachelines: 1, members: 1 */
/* last cacheline: 20 bytes */
};

Also, refactor the code accordingly and make use of the struct_size()
and flex_array_size() helpers.

This helps with the ongoing efforts to globally enable -Warray-bounds
and get us closer to being able to tighten the FORTIFY_SOURCE routines
on memcpy().

[1] https://en.wikipedia.org/wiki/Flexible_array_member
[2] https://www.kernel.org/doc/html/v5.10/process/deprecated.html#zero-length-and-one-element-arrays

Link: https://github.com/KSPP/linux/issues/79
Link: https://github.com/KSPP/linux/issues/109
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: remove the struct packet_type argument from dsa_device_ops::rcv()
Vladimir Oltean [Sat, 31 Jul 2021 14:14:32 +0000 (17:14 +0300)]
net: dsa: remove the struct packet_type argument from dsa_device_ops::rcv()

No tagging driver uses this.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonfc: hci: pass callback data param as pointer in nci_request()
Krzysztof Kozlowski [Sat, 31 Jul 2021 10:21:44 +0000 (12:21 +0200)]
nfc: hci: pass callback data param as pointer in nci_request()

The nci_request() receives a callback function and unsigned long data
argument "opt" which is passed to the callback.  Almost all of the
nci_request() callers pass pointer to a stack variable as data argument.
Only few pass scalar value (e.g. u8).

All such callbacks do not modify passed data argument and in previous
commit they were made as const.  However passing pointers via unsigned
long removes the const annotation.  The callback could simply cast
unsigned long to a pointer to writeable memory.

Use "const void *" as type of this "opt" argument to solve this and
prevent modifying the pointed contents.  This is also consistent with
generic pattern of passing data arguments - via "void *".  In few places
which pass scalar values, use casts via "unsigned long" to suppress any
warnings.

Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agocavium: switch from 'pci_' to 'dma_' API
Christophe JAILLET [Sat, 31 Jul 2021 07:10:00 +0000 (09:10 +0200)]
cavium: switch from 'pci_' to 'dma_' API

The wrappers in include/linux/pci-dma-compat.h should go away.

The patch has been generated with the coccinelle script below. It has been
hand modified to use 'dma_set_mask_and_coherent()' instead of
'pci_set_dma_mask()/pci_set_consistent_dma_mask()' when applicable.

It has been compile tested.

@@
@@
-    PCI_DMA_BIDIRECTIONAL
+    DMA_BIDIRECTIONAL

@@
@@
-    PCI_DMA_TODEVICE
+    DMA_TO_DEVICE

@@
@@
-    PCI_DMA_FROMDEVICE
+    DMA_FROM_DEVICE

@@
@@
-    PCI_DMA_NONE
+    DMA_NONE

@@
expression e1, e2, e3;
@@
-    pci_alloc_consistent(e1, e2, e3)
+    dma_alloc_coherent(&e1->dev, e2, e3, GFP_)

@@
expression e1, e2, e3;
@@
-    pci_zalloc_consistent(e1, e2, e3)
+    dma_alloc_coherent(&e1->dev, e2, e3, GFP_)

@@
expression e1, e2, e3, e4;
@@
-    pci_free_consistent(e1, e2, e3, e4)
+    dma_free_coherent(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_map_single(e1, e2, e3, e4)
+    dma_map_single(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_single(e1, e2, e3, e4)
+    dma_unmap_single(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4, e5;
@@
-    pci_map_page(e1, e2, e3, e4, e5)
+    dma_map_page(&e1->dev, e2, e3, e4, e5)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_page(e1, e2, e3, e4)
+    dma_unmap_page(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_map_sg(e1, e2, e3, e4)
+    dma_map_sg(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_unmap_sg(e1, e2, e3, e4)
+    dma_unmap_sg(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_single_for_cpu(e1, e2, e3, e4)
+    dma_sync_single_for_cpu(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_single_for_device(e1, e2, e3, e4)
+    dma_sync_single_for_device(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_sg_for_cpu(e1, e2, e3, e4)
+    dma_sync_sg_for_cpu(&e1->dev, e2, e3, e4)

@@
expression e1, e2, e3, e4;
@@
-    pci_dma_sync_sg_for_device(e1, e2, e3, e4)
+    dma_sync_sg_for_device(&e1->dev, e2, e3, e4)

@@
expression e1, e2;
@@
-    pci_dma_mapping_error(e1, e2)
+    dma_mapping_error(&e1->dev, e2)

@@
expression e1, e2;
@@
-    pci_set_dma_mask(e1, e2)
+    dma_set_mask(&e1->dev, e2)

@@
expression e1, e2;
@@
-    pci_set_consistent_dma_mask(e1, e2)
+    dma_set_coherent_mask(&e1->dev, e2)

Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet: dsa: mt7530: drop paranoid checks in .get_tag_protocol()
Vladimir Oltean [Fri, 30 Jul 2021 22:57:14 +0000 (01:57 +0300)]
net: dsa: mt7530: drop paranoid checks in .get_tag_protocol()

It is desirable to reduce the surface of DSA_TAG_PROTO_NONE as much as
we can, because we now have options for switches without hardware
support for DSA tagging, and the occurrence in the mt7530 driver is in
fact quite gratuitout and easy to remove. Since ds->ops->get_tag_protocol()
is only called for CPU ports, the checks for a CPU port in
mtk_get_tag_protocol() are redundant and can be removed.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: DENG Qingfang <dqfext@gmail.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge branch 'octeon-drr-config'
David S. Miller [Mon, 2 Aug 2021 09:47:12 +0000 (10:47 +0100)]
Merge branch 'octeon-drr-config'

Sunil Goutham says:

====================
cn10k: DWRR MTU and weights configuration

On OcteonTx2 DWRR quantum is directly configured into each of
the transmit scheduler queues. And PF/VF drivers were free to
config any value upto 2^24.

On CN10K, HW is modified, the quantum configuration at scheduler
queues is in terms of weight. And SW needs to setup a base DWRR MTU
at NIX_AF_DWRR_RPM_MTU / NIX_AF_DWRR_SDP_MTU. HW will do
'DWRR MTU * weight' to get the quantum.

This patch series addresses this HW change on CN10K silicons,
both admin function and PF/VF drivers are modified.

Also added support to program DWRR MTU via devlink params.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoocteontx2-pf: cn10k: Config DWRR weight based on MTU
Sunil Goutham [Fri, 30 Jul 2021 11:49:14 +0000 (17:19 +0530)]
octeontx2-pf: cn10k: Config DWRR weight based on MTU

Program SQ, MDQ, TL4 to TL2 transmit scheduler queues' DWRR
weight based on DWRR MTU programmed at NIX_AF_DWRR_RPM_MTU.
The DWRR MTU from admin function is retrieved via mbox.

On OcteaonTx2 silicon, admin function driver responds with DWRR
MTU as '1'. This helps to avoid silicon specific transmit
scheduler DWRR quantum/weight configuration logic.

Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoocteontx2-af: cn10k: DWRR MTU configuration
Sunil Goutham [Fri, 30 Jul 2021 11:49:13 +0000 (17:19 +0530)]
octeontx2-af: cn10k: DWRR MTU configuration

On OcteonTx2 DWRR quantum is directly configured into each of
the transmit scheduler queues. And PF/VF drivers were free to
config any value upto 2^24.

On CN10K, HW is modified, the quantum configuration at scheduler
queues is in terms of weight. And SW needs to setup a base DWRR MTU
at NIX_AF_DWRR_RPM_MTU / NIX_AF_DWRR_SDP_MTU. HW will do
'DWRR MTU * weight' to get the quantum. For LBK traffic, value
programmed into NIX_AF_DWRR_RPM_MTU register is considered as
DWRR MTU.

This patch programs a default DWRR MTU of 8192 into HW and also
provides a way to change this via devlink params.

Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests/net: remove min gso test in packet_snd
Dust Li [Fri, 30 Jul 2021 03:41:55 +0000 (11:41 +0800)]
selftests/net: remove min gso test in packet_snd

This patch removed the 'raw gso min size - 1' test which
always fails now:
./in_netns.sh ./psock_snd -v -c -g -l "${mss}"
  raw gso min size - 1 (expected to fail)
  tx: 1524
  rx: 1472
  OK

After commit 7c6d2ecbda83 ("net: be more gentle about silly
gso requests coming from user"), we relaxed the min gso_size
check in virtio_net_hdr_to_skb().
So when a packet which is smaller then the gso_size,
GSO for this packet will not be set, the packet will be
send/recv successfully.

Signed-off-by: Dust Li <dust.li@linux.alibaba.com>
Reviewed-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agobonding: 3ad: fix the concurrency between __bond_release_one() and bond_3ad_state_mac...
Yufeng Mo [Fri, 30 Jul 2021 02:19:11 +0000 (10:19 +0800)]
bonding: 3ad: fix the concurrency between __bond_release_one() and bond_3ad_state_machine_handler()

Some time ago, I reported a calltrace issue
"did not find a suitable aggregator", please see[1].
After a period of analysis and reproduction, I find
that this problem is caused by concurrency.

Before the problem occurs, the bond structure is like follows:

bond0 - slaver0(eth0) - agg0.lag_ports -> port0 - port1
                      \
                        port0
      \
        slaver1(eth1) - agg1.lag_ports -> NULL
                      \
                        port1

If we run 'ifenslave bond0 -d eth1', the process is like below:

excuting __bond_release_one()
|
bond_upper_dev_unlink()[step1]
|                       |                       |
|                       |                       bond_3ad_lacpdu_recv()
|                       |                       ->bond_3ad_rx_indication()
|                       |                       spin_lock_bh()
|                       |                       ->ad_rx_machine()
|                       |                       ->__record_pdu()[step2]
|                       |                       spin_unlock_bh()
|                       |                       |
|                       bond_3ad_state_machine_handler()
|                       spin_lock_bh()
|                       ->ad_port_selection_logic()
|                       ->try to find free aggregator[step3]
|                       ->try to find suitable aggregator[step4]
|                       ->did not find a suitable aggregator[step5]
|                       spin_unlock_bh()
|                       |
|                       |
bond_3ad_unbind_slave() |
spin_lock_bh()
spin_unlock_bh()

step1: already removed slaver1(eth1) from list, but port1 remains
step2: receive a lacpdu and update port0
step3: port0 will be removed from agg0.lag_ports. The struct is
       "agg0.lag_ports -> port1" now, and agg0 is not free. At the
   same time, slaver1/agg1 has been removed from the list by step1.
   So we can't find a free aggregator now.
step4: can't find suitable aggregator because of step2
step5: cause a calltrace since port->aggregator is NULL

To solve this concurrency problem, put bond_upper_dev_unlink()
after bond_3ad_unbind_slave(). In this way, we can invalid the port
first and skip this port in bond_3ad_state_machine_handler(). This
eliminates the situation that the slaver has been removed from the
list but the port is still valid.

[1]https://lore.kernel.org/netdev/10374.1611947473@famine/

Signed-off-by: Yufeng Mo <moyufeng@huawei.com>
Acked-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agonet_sched: refactor TC action init API
Cong Wang [Thu, 29 Jul 2021 23:12:14 +0000 (16:12 -0700)]
net_sched: refactor TC action init API

TC action ->init() API has 10 parameters, it becomes harder
to read. Some of them are just boolean and can be replaced
by flags. Similarly for the internal API tcf_action_init()
and tcf_exts_validate().

This patch converts them to flags and fold them into
the upper 16 bits of "flags", whose lower 16 bits are still
reserved for user-space. More specifically, the following
kernel flags are introduced:

TCA_ACT_FLAGS_POLICE replace 'name' in a few contexts, to
distinguish whether it is compatible with policer.

TCA_ACT_FLAGS_BIND replaces 'bind', to indicate whether
this action is bound to a filter.

TCA_ACT_FLAGS_REPLACE  replaces 'ovr' in most contexts,
means we are replacing an existing action.

TCA_ACT_FLAGS_NO_RTNL replaces 'rtnl_held' but has the
opposite meaning, because we still hold RTNL in most
cases.

The only user-space flag TCA_ACT_FLAGS_NO_PERCPU_STATS is
untouched and still stored as before.

I have tested this patch with tdc and I do not see any
failure related to this patch.

Tested-by: Vlad Buslov <vladbu@nvidia.com>
Acked-by: Jamal Hadi Salim<jhs@mojatatu.com>
Cc: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoniu: read property length only if we use it
Martin Kaiser [Thu, 29 Jul 2021 07:43:54 +0000 (09:43 +0200)]
niu: read property length only if we use it

In three places, the driver calls of_get_property and reads the property
length although the length is not used. Update the calls to not request
the length.

Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
Jakub Kicinski [Sat, 31 Jul 2021 18:23:25 +0000 (11:23 -0700)]
Merge https://git./linux/kernel/git/bpf/bpf-next

Andrii Nakryiko says:

====================
bpf-next 2021-07-30

We've added 64 non-merge commits during the last 15 day(s) which contain
a total of 83 files changed, 5027 insertions(+), 1808 deletions(-).

The main changes are:

1) BTF-guided binary data dumping libbpf API, from Alan.

2) Internal factoring out of libbpf CO-RE relocation logic, from Alexei.

3) Ambient BPF run context and cgroup storage cleanup, from Andrii.

4) Few small API additions for libbpf 1.0 effort, from Evgeniy and Hengqi.

5) bpf_program__attach_kprobe_opts() fixes in libbpf, from Jiri.

6) bpf_{get,set}sockopt() support in BPF iterators, from Martin.

7) BPF map pinning improvements in libbpf, from Martynas.

8) Improved module BTF support in libbpf and bpftool, from Quentin.

9) Bpftool cleanups and documentation improvements, from Quentin.

10) Libbpf improvements for supporting CO-RE on old kernels, from Shuyi.

11) Increased maximum cgroup storage size, from Stanislav.

12) Small fixes and improvements to BPF tests and samples, from various folks.

* https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (64 commits)
  tools: bpftool: Complete metrics list in "bpftool prog profile" doc
  tools: bpftool: Document and add bash completion for -L, -B options
  selftests/bpf: Update bpftool's consistency script for checking options
  tools: bpftool: Update and synchronise option list in doc and help msg
  tools: bpftool: Complete and synchronise attach or map types
  selftests/bpf: Check consistency between bpftool source, doc, completion
  tools: bpftool: Slightly ease bash completion updates
  unix_bpf: Fix a potential deadlock in unix_dgram_bpf_recvmsg()
  libbpf: Add btf__load_vmlinux_btf/btf__load_module_btf
  tools: bpftool: Support dumping split BTF by id
  libbpf: Add split BTF support for btf__load_from_kernel_by_id()
  tools: Replace btf__get_from_id() with btf__load_from_kernel_by_id()
  tools: Free BTF objects at various locations
  libbpf: Rename btf__get_from_id() as btf__load_from_kernel_by_id()
  libbpf: Rename btf__load() as btf__load_into_kernel()
  libbpf: Return non-null error on failures in libbpf_find_prog_btf_id()
  bpf: Emit better log message if bpf_iter ctx arg btf_id == 0
  tools/resolve_btfids: Emit warnings and patch zero id for missing symbols
  bpf: Increase supported cgroup storage value size
  libbpf: Fix race when pinning maps in parallel
  ...
====================

Link: https://lore.kernel.org/r/20210730225606.1897330-1-andrii@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Jakub Kicinski [Sat, 31 Jul 2021 16:14:46 +0000 (09:14 -0700)]
Merge git://git./linux/kernel/git/netdev/net

Conflicting commits, all resolutions pretty trivial:

drivers/bus/mhi/pci_generic.c
  5c2c85315948 ("bus: mhi: pci-generic: configurable network interface MRU")
  56f6f4c4eb2a ("bus: mhi: pci_generic: Apply no-op for wake using sideband wake boolean")

drivers/nfc/s3fwrn5/firmware.c
  a0302ff5906a ("nfc: s3fwrn5: remove unnecessary label")
  46573e3ab08f ("nfc: s3fwrn5: fix undefined parameter values in dev_err()")
  801e541c79bb ("nfc: s3fwrn5: fix undefined parameter values in dev_err()")

MAINTAINERS
  7d901a1e878a ("net: phy: add Maxlinear GPY115/21x/24x driver")
  8a7b46fa7902 ("MAINTAINERS: add Yasushi SHOJI as reviewer for the Microchip CAN BUS Analyzer Tool driver")

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'net-5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Fri, 30 Jul 2021 23:01:36 +0000 (16:01 -0700)]
Merge tag 'net-5.14-rc4' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Networking fixes for 5.14-rc4, including fixes from bpf, can, WiFi
  (mac80211) and netfilter trees.

  Current release - regressions:

   - mac80211: fix starting aggregation sessions on mesh interfaces

  Current release - new code bugs:

   - sctp: send pmtu probe only if packet loss in Search Complete state

   - bnxt_en: add missing periodic PHC overflow check

   - devlink: fix phys_port_name of virtual port and merge error

   - hns3: change the method of obtaining default ptp cycle

   - can: mcba_usb_start(): add missing urb->transfer_dma initialization

  Previous releases - regressions:

   - set true network header for ECN decapsulation

   - mlx5e: RX, avoid possible data corruption w/ relaxed ordering and
     LRO

   - phy: re-add check for PHY_BRCM_DIS_TXCRXC_NOENRGY on the BCM54811
     PHY

   - sctp: fix return value check in __sctp_rcv_asconf_lookup

  Previous releases - always broken:

   - bpf:
       - more spectre corner case fixes, introduce a BPF nospec
         instruction for mitigating Spectre v4
       - fix OOB read when printing XDP link fdinfo
       - sockmap: fix cleanup related races

   - mac80211: fix enabling 4-address mode on a sta vif after assoc

   - can:
       - raw: raw_setsockopt(): fix raw_rcv panic for sock UAF
       - j1939: j1939_session_deactivate(): clarify lifetime of session
         object, avoid UAF
       - fix number of identical memory leaks in USB drivers

   - tipc:
       - do not blindly write skb_shinfo frags when doing decryption
       - fix sleeping in tipc accept routine"

* tag 'net-5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (91 commits)
  gve: Update MAINTAINERS list
  can: esd_usb2: fix memory leak
  can: ems_usb: fix memory leak
  can: usb_8dev: fix memory leak
  can: mcba_usb_start(): add missing urb->transfer_dma initialization
  can: hi311x: fix a signedness bug in hi3110_cmd()
  MAINTAINERS: add Yasushi SHOJI as reviewer for the Microchip CAN BUS Analyzer Tool driver
  bpf: Fix leakage due to insufficient speculative store bypass mitigation
  bpf: Introduce BPF nospec instruction for mitigating Spectre v4
  sis900: Fix missing pci_disable_device() in probe and remove
  net: let flow have same hash in two directions
  nfc: nfcsim: fix use after free during module unload
  tulip: windbond-840: Fix missing pci_disable_device() in probe and remove
  sctp: fix return value check in __sctp_rcv_asconf_lookup
  nfc: s3fwrn5: fix undefined parameter values in dev_err()
  net/mlx5: Fix mlx5_vport_tbl_attr chain from u16 to u32
  net/mlx5e: Fix nullptr in mlx5e_hairpin_get_mdev()
  net/mlx5: Unload device upon firmware fatal error
  net/mlx5e: Fix page allocation failure for ptp-RQ over SF
  net/mlx5e: Fix page allocation failure for trap-RQ over SF
  ...

2 years agoMerge tag 'acpi-5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 30 Jul 2021 22:56:24 +0000 (15:56 -0700)]
Merge tag 'acpi-5.14-rc4' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fixes from Rafael Wysocki:
 "These revert a recent IRQ resources handling modification that turned
  out to be problematic, fix suspend-to-idle handling on AMD platforms
  to take upcoming systems into account properly and fix the retrieval
  of the DPTF attributes of the PCH FIVR.

  Specifics:

   - Revert recent change of the ACPI IRQ resources handling that
     attempted to improve the ACPI IRQ override selection logic, but
     introduced serious regressions on some systems (Hui Wang).

   - Fix up quirks for AMD platforms in the suspend-to-idle support code
     so as to take upcoming systems using uPEP HID AMDI007 into account
     as appropriate (Mario Limonciello).

   - Fix the code retrieving DPTF attributes of the PCH FIVR so that it
     agrees on the return data type with the ACPI control method
     evaluated for this purpose (Srinivas Pandruvada)"

* tag 'acpi-5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI: DPTF: Fix reading of attributes
  Revert "ACPI: resources: Add checks for ACPI IRQ override"
  ACPI: PM: Add support for upcoming AMD uPEP HID AMDI007

2 years agopipe: make pipe writes always wake up readers
Linus Torvalds [Fri, 30 Jul 2021 22:42:34 +0000 (15:42 -0700)]
pipe: make pipe writes always wake up readers

Since commit 1b6b26ae7053 ("pipe: fix and clarify pipe write wakeup
logic") we have sanitized the pipe write logic, and would only try to
wake up readers if they needed it.

In particular, if the pipe already had data in it before the write,
there was no point in trying to wake up a reader, since any existing
readers must have been aware of the pre-existing data already.  Doing
extraneous wakeups will only cause potential thundering herd problems.

However, it turns out that some Android libraries have misused the EPOLL
interface, and expected "edge triggered" be to "any new write will
trigger it".  Even if there was no edge in sight.

Quoting Sandeep Patil:
 "The commit 1b6b26ae7053 ('pipe: fix and clarify pipe write wakeup
  logic') changed pipe write logic to wakeup readers only if the pipe
  was empty at the time of write. However, there are libraries that
  relied upon the older behavior for notification scheme similar to
  what's described in [1]

  One such library 'realm-core'[2] is used by numerous Android
  applications. The library uses a similar notification mechanism as GNU
  Make but it never drains the pipe until it is full. When Android moved
  to v5.10 kernel, all applications using this library stopped working.

  The library has since been fixed[3] but it will be a while before all
  applications incorporate the updated library"

Our regression rule for the kernel is that if applications break from
new behavior, it's a regression, even if it was because the application
did something patently wrong.  Also note the original report [4] by
Michal Kerrisk about a test for this epoll behavior - but at that point
we didn't know of any actual broken use case.

So add the extraneous wakeup, to approximate the old behavior.

[ I say "approximate", because the exact old behavior was to do a wakeup
  not for each write(), but for each pipe buffer chunk that was filled
  in. The behavior introduced by this change is not that - this is just
  "every write will cause a wakeup, whether necessary or not", which
  seems to be sufficient for the broken library use. ]

It's worth noting that this adds the extraneous wakeup only for the
write side, while the read side still considers the "edge" to be purely
about reading enough from the pipe to allow further writes.

See commit f467a6a66419 ("pipe: fix and clarify pipe read wakeup logic")
for the pipe read case, which remains that "only wake up if the pipe was
full, and we read something from it".

Link: https://lore.kernel.org/lkml/CAHk-=wjeG0q1vgzu4iJhW5juPkTsjTYmiqiMUYAebWW+0bam6w@mail.gmail.com/
Link: https://github.com/realm/realm-core
Link: https://github.com/realm/realm-core/issues/4666
Link: https://lore.kernel.org/lkml/CAKgNAkjMBGeAwF=2MKK758BhxvW58wYTgYKB2V-gY1PwXxrH+Q@mail.gmail.com/
Link: https://lore.kernel.org/lkml/20210729222635.2937453-1-sspatil@android.com/
Reported-by: Sandeep Patil <sspatil@android.com>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoMerge branch 'tools: bpftool: update, synchronise and validate types and options'
Andrii Nakryiko [Fri, 30 Jul 2021 22:40:28 +0000 (15:40 -0700)]
Merge branch 'tools: bpftool: update, synchronise and validate types and options'

Quentin Monnet says:

====================

To work with the different program types, map types, attach types etc.
supported by eBPF, bpftool needs occasional updates to learn about the new
features supported by the kernel. When such types translate into new
keyword for the command line, updates are expected in several locations:
typically, the help message displayed from bpftool itself, the manual page,
and the bash completion file should be updated. The options used by the
different commands for bpftool should also remain synchronised at those
locations.

Several omissions have occurred in the past, and a number of types are
still missing today. This set is an attempt to improve the situation. It
brings up-to-date the lists of types or options in bpftool, and also adds a
Python script to the BPF selftests to automatically check that most of
these lists remain synchronised.

v2:
- Reformat some lines in the bash completion file.
- Do not reformat attach types, to preserve git-blame history.
- Do not call Python script from tools/testing/selftests/bpf/Makefile.
====================

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
2 years agotools: bpftool: Complete metrics list in "bpftool prog profile" doc
Quentin Monnet [Fri, 30 Jul 2021 21:54:35 +0000 (22:54 +0100)]
tools: bpftool: Complete metrics list in "bpftool prog profile" doc

Profiling programs with bpftool was extended some time ago to support
two new metrics, namely itlb_misses and dtlb_misses (misses for the
instruction/data translation lookaside buffer). Update the manual page
and bash completion accordingly.

Fixes: 450d060e8f75 ("bpftool: Add {i,d}tlb_misses support for bpftool profile")
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210730215435.7095-8-quentin@isovalent.com
2 years agotools: bpftool: Document and add bash completion for -L, -B options
Quentin Monnet [Fri, 30 Jul 2021 21:54:34 +0000 (22:54 +0100)]
tools: bpftool: Document and add bash completion for -L, -B options

The -L|--use-loader option for using loader programs when loading, or
when generating a skeleton, did not have any documentation or bash
completion. Same thing goes for -B|--base-btf, used to pass a path to a
base BTF object for split BTF such as BTF for kernel modules.

This patch documents and adds bash completion for those options.

Fixes: 75fa1777694c ("tools/bpftool: Add bpftool support for split BTF")
Fixes: d510296d331a ("bpftool: Use syscall/loader program in "prog load" and "gen skeleton" command.")
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210730215435.7095-7-quentin@isovalent.com
2 years agoselftests/bpf: Update bpftool's consistency script for checking options
Quentin Monnet [Fri, 30 Jul 2021 21:54:33 +0000 (22:54 +0100)]
selftests/bpf: Update bpftool's consistency script for checking options

Update the script responsible for checking that the different types used
at various places in bpftool are synchronised, and extend it to check
the consistency of options between the help messages in the source code
and the manual pages.

Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210730215435.7095-6-quentin@isovalent.com
2 years agotools: bpftool: Update and synchronise option list in doc and help msg
Quentin Monnet [Fri, 30 Jul 2021 21:54:32 +0000 (22:54 +0100)]
tools: bpftool: Update and synchronise option list in doc and help msg

All bpftool commands support the options for JSON output and debug from
libbpf. In addition, some commands support additional options
corresponding to specific use cases.

The list of options described in the man pages for the different
commands are not always accurate. The messages for interactive help are
mostly limited to HELP_SPEC_OPTIONS, and are even less representative of
the actual set of options supported for the commands.

Let's update the lists:

- HELP_SPEC_OPTIONS is modified to contain the "default" options (JSON
  and debug), and to be extensible (no ending curly bracket).
- All commands use HELP_SPEC_OPTIONS in their help message, and then
  complete the list with their specific options.
- The lists of options in the man pages are updated.
- The formatting of the list for bpftool.rst is adjusted to match
  formatting for the other man pages. This is for consistency, and also
  because it will be helpful in a future patch to automatically check
  that the files are synchronised.

Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210730215435.7095-5-quentin@isovalent.com
2 years agotools: bpftool: Complete and synchronise attach or map types
Quentin Monnet [Fri, 30 Jul 2021 21:54:31 +0000 (22:54 +0100)]
tools: bpftool: Complete and synchronise attach or map types

Update bpftool's list of attach type names to tell it about the latest
attach types, or the "ringbuf" map. Also update the documentation, help
messages, and bash completion when relevant.

These missing items were reported by the newly added Python script used
to help maintain consistency in bpftool.

Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210730215435.7095-4-quentin@isovalent.com