Since both tx and rx paths work with skb->vlan_tci, there's no need for
this function anymore. Switch users directly to __vlan_hwaccel_put_tag.
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Always returns the same skb it gets, so change to void.
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Acked-by: NPravin B Shelar <pshelar@nicira.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add generic RMII-Reference-Clock-Select support.

Several Micrel PHY have an RMII-Reference-Clock-Select bit to select
25 MHz or 50 MHz clock mode. Recently, support for configuring this
through device tree for KSZ8021 and KSZ8031 was added.

Generalise this support so that it can be configured for other PHY types
as well.

Note that some PHY revisions (of the same type) has this bit inverted.
This should be either configurable through a new device-tree property,
or preferably, determined based on PHY ID if possible.

Also note that this removes support for setting 25 MHz mode from board
files which was also added by the above mentioned commit 45f56cb82e45
("net/phy: micrel: Add clock support for KSZ8021/KSZ8031").
Signed-off-by: NJohan Hovold <johan@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add static driver-data field to struct phy_driver, which can be used to
store structured device-type information.
Signed-off-by: NJohan Hovold <johan@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

no callers since 3.0
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

... and do the same on the compat side of things.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Kernel-side struct msghdr is (currently) using the same layout as
userland one, but it's not a one-to-one copy - even without considering
32bit compat issues, we have msg_iov, msg_name and msg_control copied
to kernel[1].  It's fairly localized, so we get away with a few functions
where that knowledge is needed (and we could shrink that set even
more).  Pretty much everything deals with the kernel-side variant and
the few places that want userland one just use a bunch of force-casts
to paper over the differences.

The thing is, kernel-side definition of struct msghdr is *not* exposed
in include/uapi - libc doesn't see it, etc.  So we can add struct user_msghdr,
with proper annotations and let the few places that ever deal with those
beasts use it for userland pointers.  Saner typechecking aside, that will
allow to change the layout of kernel-side msghdr - e.g. replace
msg_iov/msg_iovlen there with struct iov_iter, getting rid of the need
to modify the iovec as we copy data to/from it, etc.

We could introduce kernel_msghdr instead, but that would create much more
noise - the absolute majority of the instances would need to have the
type switched to kernel_msghdr and definition of struct msghdr in
include/linux/socket.h is not going to be seen by userland anyway.

This commit just introduces user_msghdr and switches the few places that
are dealing with userland-side msghdr to it.

[1] actually, it's even trickier than that - we copy msg_control for
sendmsg, but keep the userland address on recvmsg.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

If icmp_rcv() has successfully processed the incoming ICMP datagram, we
should use consume_skb() rather than kfree_skb() because a hit on the likes
of perf -e skb:kfree_skb is not called-for.
Signed-off-by: NRick Jones <rick.jones2@hp.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

expose bpf_map_lookup_elem(), bpf_map_update_elem(), bpf_map_delete_elem()
map accessors to eBPF programs
Signed-off-by: NAlexei Starovoitov <ast@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

add new map type BPF_MAP_TYPE_ARRAY and its implementation

- optimized for fastest possible lookup()
  . in the future verifier/JIT may recognize lookup() with constant key
    and optimize it into constant pointer. Can optimize non-constant
    key into direct pointer arithmetic as well, since pointers and
    value_size are constant for the life of the eBPF program.
    In other words array_map_lookup_elem() may be 'inlined' by verifier/JIT
    while preserving concurrent access to this map from user space

- two main use cases for array type:
  . 'global' eBPF variables: array of 1 element with key=0 and value is a
    collection of 'global' variables which programs can use to keep the state
    between events
  . aggregation of tracing events into fixed set of buckets

- all array elements pre-allocated and zero initialized at init time

- key as an index in array and can only be 4 byte

- map_delete_elem() returns EINVAL, since elements cannot be deleted

- map_update_elem() replaces elements in an non-atomic way
  (for atomic updates hashtable type should be used instead)
Signed-off-by: NAlexei Starovoitov <ast@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

add new map type BPF_MAP_TYPE_HASH and its implementation

- maps are created/destroyed by userspace. Both userspace and eBPF programs
  can lookup/update/delete elements from the map

- eBPF programs can be called in_irq(), so use spin_lock_irqsave() mechanism
  for concurrent updates

- key/value are opaque range of bytes (aligned to 8 bytes)

- user space provides 3 configuration attributes via BPF syscall:
  key_size, value_size, max_entries

- map takes care of allocating/freeing key/value pairs

- map_update_elem() must fail to insert new element when max_entries
  limit is reached to make sure that eBPF programs cannot exhaust memory

- map_update_elem() replaces elements in an atomic way

- optimized for speed of lookup() which can be called multiple times from
  eBPF program which itself is triggered by high volume of events
  . in the future JIT compiler may recognize lookup() call and optimize it
    further, since key_size is constant for life of eBPF program
Signed-off-by: NAlexei Starovoitov <ast@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

the current meaning of BPF_MAP_UPDATE_ELEM syscall command is:
either update existing map element or create a new one.
Initially the plan was to add a new command to handle the case of
'create new element if it didn't exist', but 'flags' style looks
cleaner and overall diff is much smaller (more code reused), so add 'flags'
attribute to BPF_MAP_UPDATE_ELEM command with the following meaning:
 #define BPF_ANY	0 /* create new element or update existing */
 #define BPF_NOEXIST	1 /* create new element if it didn't exist */
 #define BPF_EXIST	2 /* update existing element */

bpf_update_elem(fd, key, value, BPF_NOEXIST) call can fail with EEXIST
if element already exists.

bpf_update_elem(fd, key, value, BPF_EXIST) can fail with ENOENT
if element doesn't exist.

Userspace will call it as:
int bpf_update_elem(int fd, void *key, void *value, __u64 flags)
{
    union bpf_attr attr = {
        .map_fd = fd,
        .key = ptr_to_u64(key),
        .value = ptr_to_u64(value),
        .flags = flags;
    };

    return bpf(BPF_MAP_UPDATE_ELEM, &attr, sizeof(attr));
}

First two bits of 'flags' are used to encode style of bpf_update_elem() command.
Bits 2-63 are reserved for future use.
Signed-off-by: NAlexei Starovoitov <ast@plumgrid.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use of well known RSS key increases attack surface.
Switch to a random one, using generic helper so that all
ports share a common key.

Also provide ethtool -x support to fetch RSS key
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Amir Vadai <amirv@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

RSS (Receive Side Scaling) typically uses Toeplitz hash and a 40 or 52 bytes
RSS key.

Some drivers use a constant (and well known key), some drivers use a random
key per port, making bonding setups hard to tune. Well known keys increase
attack surface, considering that number of queues is usually a power of two.

This patch provides infrastructure to help drivers doing the right thing.

netdev_rss_key_fill() should be used by drivers to initialize their RSS key,
even if they provide ethtool -X support to let user redefine the key later.

A new /proc/sys/net/core/netdev_rss_key file can be used to get the host
RSS key even for drivers not providing ethtool -x support, in case some
applications want to precisely setup flows to match some RX queues.

Tested:

myhost:~# cat /proc/sys/net/core/netdev_rss_key
11:63:99:bb:79:fb:a5:a7:07:45:b2:20:bf:02:42:2d:08:1a:dd:19:2b:6b:23:ac:56:28:9d:70:c3:ac:e8:16:4b:b7:c1:10:53:a4:78:41:36:40:74:b6:15:ca:27:44:aa:b3:4d:72

myhost:~# ethtool -x eth0
RX flow hash indirection table for eth0 with 8 RX ring(s):
    0:      0     1     2     3     4     5     6     7
RSS hash key:
11:63:99:bb:79:fb:a5:a7:07:45:b2:20:bf:02:42:2d:08:1a:dd:19:2b:6b:23:ac:56:28:9d:70:c3:ac:e8:16:4b:b7:c1:10:53:a4:78:41
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This reverts commit e5a2c899.

	Commit e5a2c899 introduced an alternative_call, arch_fast_hash2,
that selects between __jhash2 and __intel_crc4_2_hash based on the
X86_FEATURE_XMM4_2.

	Unfortunately, the alternative_call system does not appear to be
suitable for use with C functions, as register usage is not handled
properly for the called functions.  The __jhash2 function in particular
clobbers registers that are not preserved when called via
alternative_call, resulting in a panic for direct callers of
arch_fast_hash2 on older CPUs lacking sse4_2.  It is possible that
__intel_crc4_2_hash works merely by chance because it uses fewer
registers.

	This commit was suggested as the source of the problem by Jesse
Gross <jesse@nicira.com>.
Signed-off-by: NJay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In free_area_init_core(), zone->managed_pages is set to an approximate
value for lowmem, and will be adjusted when the bootmem allocator frees
pages into the buddy system.

But free_area_init_core() is also called by hotadd_new_pgdat() when
hot-adding memory.  As a result, zone->managed_pages of the newly added
node's pgdat is set to an approximate value in the very beginning.

Even if the memory on that node has node been onlined,
/sys/device/system/node/nodeXXX/meminfo has wrong value:

  hot-add node2 (memory not onlined)
  cat /sys/device/system/node/node2/meminfo
  Node 2 MemTotal:       33554432 kB
  Node 2 MemFree:               0 kB
  Node 2 MemUsed:        33554432 kB
  Node 2 Active:                0 kB

This patch fixes this problem by reset node managed pages to 0 after
hot-adding a new node.

1. Move reset_managed_pages_done from reset_node_managed_pages() to
   reset_all_zones_managed_pages()
2. Make reset_node_managed_pages() non-static
3. Call reset_node_managed_pages() in hotadd_new_pgdat() after pgdat
   is initialized
Signed-off-by: NTang Chen <tangchen@cn.fujitsu.com>
Signed-off-by: NYasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: <stable@vger.kernel.org>	[3.16+]
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Before describing bugs itself, I first explain definition of freepage.

 1. pages on buddy list are counted as freepage.
 2. pages on isolate migratetype buddy list are *not* counted as freepage.
 3. pages on cma buddy list are counted as CMA freepage, too.

Now, I describe problems and related patch.

Patch 1: There is race conditions on getting pageblock migratetype that
it results in misplacement of freepages on buddy list, incorrect
freepage count and un-availability of freepage.

Patch 2: Freepages on pcp list could have stale cached information to
determine migratetype of buddy list to go.  This causes misplacement of
freepages on buddy list and incorrect freepage count.

Patch 4: Merging between freepages on different migratetype of
pageblocks will cause freepages accouting problem.  This patch fixes it.

Without patchset [3], above problem doesn't happens on my CMA allocation
test, because CMA reserved pages aren't used at all.  So there is no
chance for above race.

With patchset [3], I did simple CMA allocation test and get below
result:

 - Virtual machine, 4 cpus, 1024 MB memory, 256 MB CMA reservation
 - run kernel build (make -j16) on background
 - 30 times CMA allocation(8MB * 30 = 240MB) attempts in 5 sec interval
 - Result: more than 5000 freepage count are missed

With patchset [3] and this patchset, I found that no freepage count are
missed so that I conclude that problems are solved.

On my simple memory offlining test, these problems also occur on that
environment, too.

This patch (of 4):

There are two paths to reach core free function of buddy allocator,
__free_one_page(), one is free_one_page()->__free_one_page() and the
other is free_hot_cold_page()->free_pcppages_bulk()->__free_one_page().
Each paths has race condition causing serious problems.  At first, this
patch is focused on first type of freepath.  And then, following patch
will solve the problem in second type of freepath.

In the first type of freepath, we got migratetype of freeing page
without holding the zone lock, so it could be racy.  There are two cases
of this race.

 1. pages are added to isolate buddy list after restoring orignal
    migratetype

    CPU1                                   CPU2

    get migratetype => return MIGRATE_ISOLATE
    call free_one_page() with MIGRATE_ISOLATE

                                grab the zone lock
                                unisolate pageblock
                                release the zone lock

    grab the zone lock
    call __free_one_page() with MIGRATE_ISOLATE
    freepage go into isolate buddy list,
    although pageblock is already unisolated

This may cause two problems.  One is that we can't use this page anymore
until next isolation attempt of this pageblock, because freepage is on
isolate buddy list.  The other is that freepage accouting could be wrong
due to merging between different buddy list.  Freepages on isolate buddy
list aren't counted as freepage, but ones on normal buddy list are
counted as freepage.  If merge happens, buddy freepage on normal buddy
list is inevitably moved to isolate buddy list without any consideration
of freepage accouting so it could be incorrect.

 2. pages are added to normal buddy list while pageblock is isolated.
    It is similar with above case.

This also may cause two problems.  One is that we can't keep these
freepages from being allocated.  Although this pageblock is isolated,
freepage would be added to normal buddy list so that it could be
allocated without any restriction.  And the other problem is same as
case 1, that it, incorrect freepage accouting.

This race condition would be prevented by checking migratetype again
with holding the zone lock.  Because it is somewhat heavy operation and
it isn't needed in common case, we want to avoid rechecking as much as
possible.  So this patch introduce new variable, nr_isolate_pageblock in
struct zone to check if there is isolated pageblock.  With this, we can
avoid to re-check migratetype in common case and do it only if there is
isolated pageblock or migratetype is MIGRATE_ISOLATE.  This solve above
mentioned problems.

Changes from v3:
Add one more check in free_one_page() that checks whether migratetype is
MIGRATE_ISOLATE or not. Without this, abovementioned case 1 could happens.
Signed-off-by: NJoonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: NMinchan Kim <minchan@kernel.org>
Acked-by: NMichal Nazarewicz <mina86@mina86.com>
Acked-by: NVlastimil Babka <vbabka@suse.cz>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Laura Abbott <lauraa@codeaurora.org>
Cc: Heesub Shin <heesub.shin@samsung.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Ritesh Harjani <ritesh.list@gmail.com>
Cc: Gioh Kim <gioh.kim@lge.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Reallocation is only required for shrinking and expanding and both rely
on a mutex for synchronization and callers of rhashtable_init() are in
non atomic context. Therefore, no reason to continue passing allocation
hints through the API.

Instead, use GFP_KERNEL and add __GFP_NOWARN | __GFP_NORETRY to allow
for silent fall back to vzalloc() without the OOM killer jumping in as
pointed out by Eric Dumazet and Eric W. Biederman.
Signed-off-by: NThomas Graf <tgraf@suug.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We now allow up to 126 VFs. Note though that certain firmware
versions only allow up to 80 VFs. Moreover, old HCAs only support 64 VFs.
In these cases, we limit the maximum number of VFs to 64.
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Previously, the driver queried the firmware in order to get the number
of supported EQs. Under SRIOV, since this was done before the driver
notified the firmware how many VFs it actually needs, the firmware had
to take into account a worst case scenario and always allocated four EQs
per VF, where one was used for events while the others were used for completions.

Now, when the firmware supports the asymmetric allocation scheme, denoted
by exposing num_sys_eqs > 0 (--> MLX4_DEV_CAP_FLAG2_SYS_EQS), we use the
QUERY_FUNC command to query the firmware before enabling SRIOV. Thus we
can get more EQs and MSI-X vectors per function.

Moreover, when running in the new firmware/driver mode, the limitation
that the number of EQs should be a power of two is lifted.
Signed-off-by: NJack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently mutex_is_held can only test locks in the that are global
since it takes no arguments.  This prevents rhashtable from being
used in places where locks are lock, e.g., per-namespace locks.

This patch adds a parent field to mutex_is_held and rhashtable_params
so that local locks can be used (and tested).
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The rhashtable function mutex_is_held is only used when PROVE_LOCKING
is enabled.  This patch makes the mutex_is_held field in rhashtable
optional depending on PROVE_LOCKING.
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Instead of calling fou and gue functions directly from ip_tunnel
use ops for these that were previously registered. This patch adds the
logic to add and remove encapsulation operations for ip_tunnel,
and modified fou (and gue) to register with ip_tunnels.

This patch also addresses a circular dependency between ip_tunnel
and fou that was causing link errors when CONFIG_NET_IP_TUNNEL=y
and CONFIG_NET_FOU=m. References to fou an gue have been removed from
ip_tunnel.c
Reported-by: NRandy Dunlap <rdunlap@infradead.org>
Signed-off-by: NTom Herbert <therbert@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the normal kernel debugging mechanism which also
enables dynamic_debug at the same time.

Other miscellanea:

o Remove sysctl for irda_debug
o Remove function tracing like uses (use ftrace instead)
o Coalesce formats
o Realign arguments
o Remove unnecessary OOM messages
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add helper macro for PHY drivers which do not do anything special in
module init/exit. This will allow us to eliminate a lot of boilerplate
code.
Signed-off-by: NJohan Hovold <johan@kernel.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove the two functions which are now dead code.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This patch implements __dev_alloc_pages and __dev_alloc_page.  These are
meant to replace the __skb_alloc_pages and __skb_alloc_page functions.  The
reason for doing this is that it occurred to me that __skb_alloc_page is
supposed to be passed an sk_buff pointer, but it is NULL in all cases where
it is used.  Worse is that in the case of ixgbe it is passed NULL via the
sk_buff pointer in the rx_buffer info structure which means the compiler is
not correctly stripping it out.

The naming for these functions is based on dev_alloc_skb and __dev_alloc_skb.
There was originally a netdev_alloc_page, however that was passed a
net_device pointer and this function is not so I thought it best to follow
that naming scheme since that is the same difference between dev_alloc_skb
and netdev_alloc_skb.

In the case of anything greater than order 0 it is assumed that we want a
compound page so __GFP_COMP is set for all allocations as we expect a
compound page when assigning a page frag.

The other change in this patch is to exploit the behaviors of the page
allocator in how it handles flags.  So for example we can always set
__GFP_COMP and __GFP_MEMALLOC since they are ignored if they are not
applicable or are overridden by another flag.
Signed-off-by: NAlexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

And use the more common mechanisms directly.

Other miscellanea:

o Coalesce formats
o Add missing newlines
o Realign arguments
o Remove unnecessary OOM message logging as
  there's a generic stack dump already on OOM.
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

vlan was the only user of netif_copy_real_num_queues(),
but it no longer calls it after
commit 4af429d2 ("vlan: lockless transmit path").
So we can just remove it.

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

These are the same as net_<level>_ratelimited, so
use the more common style in the macro definition.
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently there are only three neigh tables in the whole kernel:
arp table, ndisc table and decnet neigh table. What's more,
we don't support registering multiple tables per family.
Therefore we can just make these tables statically built-in.

Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: NCong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use the more common dynamic_debug capable net_dbg_ratelimited
and remove the LIMIT_NETDEBUG macro.

All messages are still ratelimited.

Some KERN_<LEVEL> uses are changed to KERN_DEBUG.

This may have some negative impact on messages that were
emitted at KERN_INFO that are not not enabled at all unless
DEBUG is defined or dynamic_debug is enabled.  Even so,
these messages are now _not_ emitted by default.

This also eliminates the use of the net_msg_warn sysctl
"/proc/sys/net/core/warnings".  For backward compatibility,
the sysctl is not removed, but it has no function.  The extern
declaration of net_msg_warn is removed from sock.h and made
static in net/core/sysctl_net_core.c

Miscellanea:

o Update the sysctl documentation
o Remove the embedded uses of pr_fmt
o Coalesce format fragments
o Realign arguments
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

When processing received traffic, pass CHECKSUM_COMPLETE status to the
stack, with calculated checksum for non TCP/UDP packets (such
as GRE or ICMP).

Although the stack expects checksum which doesn't include the pseudo
header, the HW adds it. To address that, we are subtracting the pseudo
header checksum from the checksum value provided by the HW.

In the IPv6 case, we also compute/add the IP header checksum which
is not added by the HW for such packets.

Cc: Jerry Chu <hkchu@google.com>
Signed-off-by: NShani Michaeli <shanim@mellanox.com>
Signed-off-by: NMatan Barak <matanb@mellanox.com>
Signed-off-by: NOr Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Alternative to RPS/RFS is to use hardware support for multiple
queues.

Then split a set of million of sockets into worker threads, each
one using epoll() to manage events on its own socket pool.

Ideally, we want one thread per RX/TX queue/cpu, but we have no way to
know after accept() or connect() on which queue/cpu a socket is managed.

We normally use one cpu per RX queue (IRQ smp_affinity being properly
set), so remembering on socket structure which cpu delivered last packet
is enough to solve the problem.

After accept(), connect(), or even file descriptor passing around
processes, applications can use :

 int cpu;
 socklen_t len = sizeof(cpu);

 getsockopt(fd, SOL_SOCKET, SO_INCOMING_CPU, &cpu, &len);

And use this information to put the socket into the right silo
for optimal performance, as all networking stack should run
on the appropriate cpu, without need to send IPI (RPS/RFS).
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

On a !PREEMPT kernel, attempting to use trace-cmd results in a soft
lockup:

 # trace-cmd record -e raw_syscalls:* -F false
 NMI watchdog: BUG: soft lockup - CPU#0 stuck for 22s! [trace-cmd:61]
 ...
 Call Trace:
  [<ffffffff8105b580>] ? __wake_up_common+0x90/0x90
  [<ffffffff81092e25>] wait_on_pipe+0x35/0x40
  [<ffffffff810936e3>] tracing_buffers_splice_read+0x2e3/0x3c0
  [<ffffffff81093300>] ? tracing_stats_read+0x2a0/0x2a0
  [<ffffffff812d10ab>] ? _raw_spin_unlock+0x2b/0x40
  [<ffffffff810dc87b>] ? do_read_fault+0x21b/0x290
  [<ffffffff810de56a>] ? handle_mm_fault+0x2ba/0xbd0
  [<ffffffff81095c80>] ? trace_event_buffer_lock_reserve+0x40/0x80
  [<ffffffff810951e2>] ? trace_buffer_lock_reserve+0x22/0x60
  [<ffffffff81095c80>] ? trace_event_buffer_lock_reserve+0x40/0x80
  [<ffffffff8112415d>] do_splice_to+0x6d/0x90
  [<ffffffff81126971>] SyS_splice+0x7c1/0x800
  [<ffffffff812d1edd>] tracesys_phase2+0xd3/0xd8

The problem is this: tracing_buffers_splice_read() calls
ring_buffer_wait() to wait for data in the ring buffers.  The buffers
are not empty so ring_buffer_wait() returns immediately.  But
tracing_buffers_splice_read() calls ring_buffer_read_page() with full=1,
meaning it only wants to read a full page.  When the full page is not
available, tracing_buffers_splice_read() tries to wait again with
ring_buffer_wait(), which again returns immediately, and so on.

Fix this by adding a "full" argument to ring_buffer_wait() which will
make ring_buffer_wait() wait until the writer has left the reader's
page, i.e.  until full-page reads will succeed.

Link: http://lkml.kernel.org/r/1415645194-25379-1-git-send-email-rabin@rab.in

Cc: stable@vger.kernel.org # 3.16+
Fixes: b1169cc6 ("tracing: Remove mock up poll wait function")
Signed-off-by: NRabin Vincent <rabin@rab.in>
Signed-off-by: NSteven Rostedt <rostedt@goodmis.org>

When doing GRO processing for UDP tunnels, we never add
SKB_GSO_UDP_TUNNEL to gso_type - only the type of the inner protocol
is added (such as SKB_GSO_TCPV4). The result is that if the packet is
later resegmented we will do GSO but not treat it as a tunnel. This
results in UDP fragmentation of the outer header instead of (i.e.) TCP
segmentation of the inner header as was originally on the wire.
Signed-off-by: NJesse Gross <jesse@nicira.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This ways drivers like cxgb4 don't need to do ugly relative includes.
Reported-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Tuning coalescing parameters on NIC can be really hard.

Servers can handle both bulk and RPC like traffic, with conflicting
goals : bulk flows want as big GRO packets as possible, RPC want minimal
latencies.

To reach big GRO packets on 10Gbe NIC, one can use :

ethtool -C eth0 rx-usecs 4 rx-frames 44

But this penalizes rpc sessions, with an increase of latencies, up to
50% in some cases, as NICs generally do not force an interrupt when
a packet with TCP Push flag is received.

Some NICs do not have an absolute timer, only a timer rearmed for every
incoming packet.

This patch uses a different strategy : Let GRO stack decides what do do,
based on traffic pattern.

Packets with Push flag wont be delayed.
Packets without Push flag might be held in GRO engine, if we keep
receiving data.

This new mechanism is off by default, and shall be enabled by setting
/sys/class/net/ethX/gro_flush_timeout to a value in nanosecond.

To fully enable this mechanism, drivers should use napi_complete_done()
instead of napi_complete().

Tested:
 Ran 200 netperf TCP_STREAM from A to B (10Gbe mlx4 link, 8 RX queues)

Without this feature, we send back about 305,000 ACK per second.

GRO aggregation ratio is low (811/305 = 2.65 segments per GRO packet)

Setting a timer of 2000 nsec is enough to increase GRO packet sizes
and reduce number of ACK packets. (811/19.2 = 42)

Receiver performs less calls to upper stacks, less wakes up.
This also reduces cpu usage on the sender, as it receives less ACK
packets.

Note that reducing number of wakes up increases cpu efficiency, but can
decrease QPS, as applications wont have the chance to warmup cpu caches
doing a partial read of RPC requests/answers if they fit in one skb.

B:~# sar -n DEV 1 10 | grep eth0 | tail -1
Average:         eth0 811269.80 305732.30 1199462.57  19705.72      0.00
0.00      0.50

B:~# echo 2000 >/sys/class/net/eth0/gro_flush_timeout

B:~# sar -n DEV 1 10 | grep eth0 | tail -1
Average:         eth0 811577.30  19230.80 1199916.51   1239.80      0.00
0.00      0.50
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Babel uses rt_proto 42. Add to userspace visible header file.
Signed-off-by: NDave Taht <dave.taht@bufferbloat.net>
Signed-off-by: NStephen Hemminger <stephen@networkplumber.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

All interrupts coming from MUIC were ignored because interrupt source
register was masked.

The Maxim 77693 has a "interrupt source" - a separate register and interrupts
which give information about PMIC block triggering the individual
interrupt (charger, topsys, MUIC, flash LED).

By default bootloader could initialize this register to "mask all"
value. In such case (observed on Trats2 board) MUIC interrupts won't be
generated regardless of their mask status. Regmap irq chip was unmasking
individual MUIC interrupts but the source was masked

Before introducing regmap irq chip this interrupt source was unmasked,
read and acked. Reading and acking is not necessary but unmasking is.

Fixes: 342d669c ("mfd: max77693: Handle IRQs using regmap")

Cc: <stable@vger.kernel.org>
Signed-off-by: NKrzysztof Kozlowski <k.kozlowski@samsung.com>
Reviewed-by: NChanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: NLee Jones <lee.jones@linaro.org>