Raw sockets with hdrincl enabled can insert ipv6 extension headers
right into the data stream. In case we need to fragment those packets,
we reparse the options header to find the place where we can insert
the fragment header. If the extension headers exceed the link's MTU we
actually cannot make progress in such a case.
Instead of ending up in broken arithmetic or rounding towards 0 and
entering an endless loop in ip6_fragment, just prevent those cases by
aborting early and signal -EMSGSIZE to user space.
This is the second version of the patch which doesn't use the
overflow_usub function, which got reverted for now.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Linus dislikes these changes. To not hold up the net-merge let's revert
it for now and fix the bug like Linus suggested.
This reverts commit ec3661b422, reversing
changes made to c80dbe0461.
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Or Gerlitz says:
====================
Mellanox mlx4 driver fixes for 4.3-rc7
Jack's fix is for a regression introduced in 4.3-rc1
Carol's fix addresses an issue which exists for while and
turns to beat us hard on PPC, please queue for -stable.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
When doing memcpy/memset of EQEs, we should use sizeof struct
mlx4_eqe as the base size and not caps.eqe_size which could be bigger.
If caps.eqe_size is bigger than the struct mlx4_eqe then we corrupt
data in the master context.
When using a 64 byte stride, the memcpy copied over 63 bytes to the
slave_eq structure. This resulted in copying over the entire eqe of
interest, including its ownership bit -- and also 31 bytes of garbage
into the next WQE in the slave EQ -- which did NOT include the ownership
bit (and therefore had no impact).
However, once the stride is increased to 128, we are overwriting the
ownership bits of *three* eqes in the slave_eq struct. This results
in an incorrect ownership bit for those eqes, which causes the eq to
seem to be full. The issue therefore surfaced only once 128-byte EQEs
started being used in SRIOV and (overarchitectures that have 128/256
byte cache-lines such as PPC) - e.g after commit 77507aa249
"net/mlx4_core: Enable CQE/EQE stride support".
Fixes: 08ff32352d ('mlx4: 64-byte CQE/EQE support')
Signed-off-by: Carol L Soto <clsoto@linux.vnet.ibm.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We do not set the ins_vlan field to zero when no vlan id is present in the packet.
Since WQEs in the TX ring are not zeroed out between uses, this oversight
could result in having vlan flags present in the WQE ctrl segment when no
vlan is preset.
Fixes: e38af4faf0 ('net/mlx4_en: Add support for hardware accelerated 802.1ad vlan')
Reported-by: Gideon Naim <gideonn@mellanox.com>
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 2751c9882b ("vhost: cross-endian
support for legacy devices") introduced a minor regression: even with
cross-endian disabled, and even on LE host, vhost_is_little_endian is
checking is_le flag so there's always a branch.
To fix, simply check virtio_legacy_is_little_endian first.
Cc: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define aarch64 specific registers for building bpf samples correctly.
Signed-off-by: Yang Shi <yang.shi@linaro.org>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
During Tx cleanup it's still possible for the descriptor data to be
read ahead of the descriptor index. A memory barrier is required between
the read of the descriptor index and the start of the Tx cleanup loop.
This allows a change to a lighter-weight barrier in the Tx transmit
routine just before updating the current descriptor index.
Since the memory barrier does result in extra overhead on arm64, keep
the previous change to not chase the current descriptor value. This
prevents the execution of the barrier for each loop performed.
Suggested-by: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Either of pskb_pull() or pskb_trim() may fail under low memory conditions.
If rds_tcp_data_recv() ignores such failures, the application will
receive corrupted data because the skb has not been correctly
carved to the RDS datagram size.
Avoid this by handling pskb_pull/pskb_trim failure in the same
manner as the skb_clone failure: bail out of rds_tcp_data_recv(), and
retry via the deferred call to rds_send_worker() that gets set up on
ENOMEM from rds_tcp_read_sock()
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Forcedeth currently uses disable_irq_lockdep and enable_irq_lockdep, which in
some configurations simply calls local_irq_disable. This causes errant warnings
in the netpoll path as in netpoll_send_skb_on_dev, where we disable irqs using
local_irq_save, leading to the following warning:
WARNING: at net/core/netpoll.c:352 netpoll_send_skb_on_dev+0x243/0x250() (Not
tainted)
Hardware name:
netpoll_send_skb_on_dev(): eth0 enabled interrupts in poll
(nv_start_xmit_optimized+0x0/0x860 [forcedeth])
Modules linked in: netconsole(+) configfs ipv6 iptable_filter ip_tables ppdev
parport_pc parport sg microcode serio_raw edac_core edac_mce_amd k8temp
snd_hda_codec_realtek snd_hda_codec_generic forcedeth snd_hda_intel
snd_hda_codec snd_hwdep snd_seq snd_seq_device snd_pcm snd_timer snd soundcore
snd_page_alloc i2c_nforce2 i2c_core shpchp ext4 jbd2 mbcache sr_mod cdrom sd_mod
crc_t10dif pata_amd ata_generic pata_acpi sata_nv dm_mirror dm_region_hash
dm_log dm_mod [last unloaded: scsi_wait_scan]
Pid: 1940, comm: modprobe Not tainted 2.6.32-573.7.1.el6.x86_64.debug #1
Call Trace:
[<ffffffff8107bbc1>] ? warn_slowpath_common+0x91/0xe0
[<ffffffff8107bcc6>] ? warn_slowpath_fmt+0x46/0x60
[<ffffffffa00fe5b0>] ? nv_start_xmit_optimized+0x0/0x860 [forcedeth]
[<ffffffff814b3593>] ? netpoll_send_skb_on_dev+0x243/0x250
[<ffffffff814b37c9>] ? netpoll_send_udp+0x229/0x270
[<ffffffffa02e3299>] ? write_msg+0x39/0x110 [netconsole]
[<ffffffffa02e331b>] ? write_msg+0xbb/0x110 [netconsole]
[<ffffffff8107bd55>] ? __call_console_drivers+0x75/0x90
[<ffffffff8107bdba>] ? _call_console_drivers+0x4a/0x80
[<ffffffff8107c445>] ? release_console_sem+0xe5/0x250
[<ffffffff8107d200>] ? register_console+0x190/0x3e0
[<ffffffffa02e71a6>] ? init_netconsole+0x1a6/0x216 [netconsole]
[<ffffffffa02e7000>] ? init_netconsole+0x0/0x216 [netconsole]
[<ffffffff810020d0>] ? do_one_initcall+0xc0/0x280
[<ffffffff810d4933>] ? sys_init_module+0xe3/0x260
[<ffffffff8100b0d2>] ? system_call_fastpath+0x16/0x1b
---[ end trace f349c7af88e6a6d5 ]---
console [netcon0] enabled
netconsole: network logging started
Fix it by modifying the forcedeth code to use
disable_irq_nosync_lockdep_irqsavedisable_irq_nosync_lockdep_irqsave instead,
which saves and restores irq state properly. This also saves us a little code
in the process
Tested by the reporter, with successful restuls
Patch applies to the head of the net tree
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: "David S. Miller" <davem@davemloft.net>
Reported-by: Vasily Averin <vvs@sw.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
nf_ct_frag6_gather() makes a clone of each skb passed to it, and if the
reassembly is successful, expects the caller to free all of the original
skbs using nf_ct_frag6_consume_orig(). This call was previously missing,
meaning that the original fragments were never freed (with the exception
of the last fragment to arrive).
Fix this by ensuring that all original fragments except for the last
fragment are freed via nf_ct_frag6_consume_orig(). The last fragment
will be morphed into the head, so it must not be freed yet. Furthermore,
retain the ->next pointer for the head after skb_morph().
Fixes: 7f8a436eaa ("openvswitch: Add conntrack action")
Reported-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is needed in openvswitch to fix an skb leak in the next patch.
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If ip_defrag() returns an error other than -EINPROGRESS, then the skb is
freed. When handle_fragments() passes this back up to
do_execute_actions(), it will be freed again. Prevent this double free
by never freeing the skb in do_execute_actions() for errors returned by
ovs_ct_execute. Always free it in ovs_ct_execute() error paths instead.
Fixes: 7f8a436eaa ("openvswitch: Add conntrack action")
Reported-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Joe Stringer <joestringer@nicira.com>
Acked-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We were computing the child index in cases where the key value we were
looking for was actually less than the base key of the tnode. As a result
we were getting incorrect index values that would cause us to skip over
some children.
To fix this I have added a test that will force us to use child index 0 if
the key we are looking for is less than the key of the current tnode.
Fixes: 8be33e955c ("fib_trie: Fib walk rcu should take a tnode and key instead of a trie and a leaf")
Reported-by: Brian Rak <brak@gameservers.com>
Signed-off-by: Alexander Duyck <aduyck@mirantis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Julia Lawall says:
====================
add missing of_node_put
The various for_each device_node iterators performs an of_node_get on each
iteration, so a break out of the loop requires an of_node_put.
The complete semantic patch that fixes this problem is
(http://coccinelle.lip6.fr):
// <smpl>
@r@
local idexpression n;
expression e1,e2;
iterator name for_each_node_by_name, for_each_node_by_type,
for_each_compatible_node, for_each_matching_node,
for_each_matching_node_and_match, for_each_child_of_node,
for_each_available_child_of_node, for_each_node_with_property;
iterator i;
statement S;
expression list [n1] es;
@@
(
(
for_each_node_by_name(n,e1) S
|
for_each_node_by_type(n,e1) S
|
for_each_compatible_node(n,e1,e2) S
|
for_each_matching_node(n,e1) S
|
for_each_matching_node_and_match(n,e1,e2) S
|
for_each_child_of_node(e1,n) S
|
for_each_available_child_of_node(e1,n) S
|
for_each_node_with_property(n,e1) S
)
&
i(es,n,...) S
)
@@
local idexpression r.n;
iterator r.i;
expression e;
expression list [r.n1] es;
@@
i(es,n,...) {
...
(
of_node_put(n);
|
e = n
|
return n;
|
+ of_node_put(n);
? return ...;
)
...
}
@@
local idexpression r.n;
iterator r.i;
expression e;
expression list [r.n1] es;
@@
i(es,n,...) {
...
(
of_node_put(n);
|
e = n
|
+ of_node_put(n);
? break;
)
...
}
... when != n
@@
local idexpression r.n;
iterator r.i;
expression e;
identifier l;
expression list [r.n1] es;
@@
i(es,n,...) {
...
(
of_node_put(n);
|
e = n
|
+ of_node_put(n);
? goto l;
)
...
}
...
l: ... when != n// </smpl>
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
for_each_available_child_of_node performs an of_node_get on each iteration, so
a break out of the loop requires an of_node_put.
A simplified version of the semantic patch that fixes this problem is as
follows (http://coccinelle.lip6.fr):
// <smpl>
@@
expression root,e;
local idexpression child;
@@
for_each_available_child_of_node(root, child) {
... when != of_node_put(child)
when != e = child
(
return child;
|
+ of_node_put(child);
? return ...;
)
...
}
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
for_each_compatible_node performs an of_node_get on each iteration, so
a break out of the loop requires an of_node_put.
A simplified version of the semantic patch that fixes this problem is as
follows (http://coccinelle.lip6.fr):
// <smpl>
@@
expression e;
local idexpression n;
@@
for_each_compatible_node(n,...) {
... when != of_node_put(n)
when != e = n
(
return n;
|
+ of_node_put(n);
? return ...;
)
...
}
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
for_each_available_child_of_node performs an of_node_get on each iteration, so
a break out of the loop requires an of_node_put.
A simplified version of the semantic patch that fixes this problem is as
follows (http://coccinelle.lip6.fr):
// <smpl>
@@
expression root,e;
local idexpression child;
@@
for_each_available_child_of_node(root, child) {
... when != of_node_put(child)
when != e = child
(
return child;
|
+ of_node_put(child);
? return ...;
)
...
}
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
for_each_available_child_of_node performs an of_node_get on each iteration, so
a break out of the loop requires an of_node_put.
A simplified version of the semantic patch that fixes this problem is as
follows (http://coccinelle.lip6.fr):
// <smpl>
@@
local idexpression r.n;
expression r,e;
@@
for_each_available_child_of_node(r,n) {
...
(
of_node_put(n);
|
e = n
|
+ of_node_put(n);
? break;
)
...
}
... when != n
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
for_each_child_of_node performs an of_node_get on each iteration, so
a break out of the loop requires an of_node_put.
A simplified version of the semantic patch that fixes this problem is as
follows (http://coccinelle.lip6.fr):
// <smpl>
@@
local idexpression r.n;
expression r,e;
@@
for_each_child_of_node(r,n) {
...
(
of_node_put(n);
|
e = n
|
+ of_node_put(n);
? break;
)
...
}
... when != n
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
for_each_child_of_node performs an of_node_get on each iteration, so
a break out of the loop requires an of_node_put.
A simplified version of the semantic patch that fixes this problem is as
follows (http://coccinelle.lip6.fr):
// <smpl>
@@
local idexpression r.n;
expression r,e;
@@
for_each_child_of_node(r,n) {
...
(
of_node_put(n);
|
e = n
|
+ of_node_put(n);
? break;
)
...
}
... when != n
// </smpl>
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
gre_gso_segment() chokes if SIT frames were aggregated by GRO engine.
Fixes: 61c1db7fae ("ipv6: sit: add GSO/TSO support")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sergei Shtylyov says:
====================
sh_eth: RX buffer alignment fixes
Here's a set of 2 patches against DaveM's 'net.git' repo which are the
fixes to the RX buffer size calculation.
[1/2] sh_eth: fix RX buffer size alignment
[2/2] sh_eth: fix RX buffer size calculation
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The RX buffer size calulation failed to account for the length granularity
(which is now 32 bytes)...
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Both Renesas R-Car and RZ/A1 manuals state that RX buffer length must be
a multiple of 32 bytes, while the driver only uses 16 byte granularity...
Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Claudiu Manoil says:
====================
gianfar: Misc. fixes and updates
Various fixes for some older issues, including having a
MAINTAINERS entry for this driver.
I'd recommend applying them on top of net, thanks.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The Rx BSY error interrupt indicates that a frame was
received and discarded due to lack of buffers, so it's
a rx ring overflow condition and has nothing to do with
with bad rx packets. Use the right counter.
BSY conditions happen when the SoC is under performance
stress. Doing *more* work in stress situations by trying
to schedule NAPI is not a good idea as the stressed system
becomes still more stressed. The Rx interrupt is already
at work making sure the NAPI is scheduled.
So calling gfar_receive() here does not help. This issue
was present since day 1.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Under one unusual circumstance it's possible to wrongly set
FILREN without enabling PRSDEP as well in the RCTRL register,
against the hardware specifications. With the default config
this does not happen because the default Rx offloads (Rx csum
and Rx VLAN) properly enable PRSDEP. But if anyone disables
all these offloads (via ethtool), we get a wrong configuration
were the Rx flow classification and hashing, and other Filer
based features (e.g. wake-on-filer interrupt) won't work.
This patch fixes the issue.
Also, account for Rx FCB insertion which happens every time
PRSDEP is set.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
RQFCR_AND is duplicated.
Add missing space as well.
Signed-off-by: Claudiu Manoil <claudiu.manoil@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
David Daney says:
====================
net: thunderx: Support pass-2 revision hardware.
With the availability of a new revision of the ThunderX NIC hardware a
few changes to the driver are required. With these, the driver works
on all currently available hardware revisions.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Add support for ThunderX pass2 CPI and MPI configuration changes.
MPI_ALG is not enabled i.e MCAM parsing is disabled.
Signed-off-by: Thanneeru Srinivasulu <tsrinivasulu@caviumnetworks.com>
Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The test for pass-1 silicon was incorrect, it should be for all
revisions less than 8. Also the revision is already present in the
pci_dev, so there is no need to read and keep a private copy.
Remove rev_id and code to read it from struct nicpf. Create new
static inline function pass1_silicon() to be used to testing the
silicon version. Use pass1_silicon() for revision checks, this will
be more widely used in follow on patches.
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In some silicon revisions, the soft reset clobbers PCI config space,
so quit doing the reset.
Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can't rely on PPPOX_ZOMBIE to decide whether to clear po->pppoe_dev.
PPPOX_ZOMBIE can be set by pppoe_disc_rcv() even when po->pppoe_dev is
NULL. So we have no guarantee that (sk->sk_state & PPPOX_ZOMBIE) implies
(po->pppoe_dev != NULL).
Since we're releasing a PPPoE socket, we want to release the pppoe_dev
if it exists and reset sk_state to PPPOX_DEAD, no matter the previous
value of sk_state. So we can just check for po->pppoe_dev and avoid any
assumption on sk->sk_state.
Fixes: 2b018d57ff ("pppoe: drop PPPOX_ZOMBIEs in pppoe_release")
Signed-off-by: Guillaume Nault <g.nault@alphalink.fr>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code currently uses the lightweight dma_wmb barrier before updating
the current descriptor count. Under heavy load, the Tx cleanup routine
was seeing the updated current descriptor count before the updated
descriptor information. As a result, the Tx descriptor was being cleaned
up before it was used because it was not "owned" by the hardware yet,
resulting in a Tx queue hang.
Using the wmb barrier insures that the descriptor is updated before the
descriptor counter preventing the Tx queue hang. For extra insurance,
the Tx cleanup routine is changed to grab the current decriptor count on
entry and uses that initial value in the processing loop rather than
trying to chase the current value.
Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>
Tested-by: Christoffer Dall <christoffer.dall@linaro.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Very rarely, the KSZ9031 will appear to complete autonegotiation, but
will drop all traffic afterwards. When this happens, the idle error
count will read 0xFF after autonegotiation completes. Reset the PHY
when in that state.
Signed-off-by: Nathan Sullivan <nathan.sullivan@ni.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hannes Frederic Sowa says:
====================
overflow-arith: begin to add support for overflow builtins functions
I add a new header, linux/overflow-arith.h, as the central place to add
overflow and wrap-around checking functions. The reason I am doing so
is that it can make use of compiler supported builtin functions which
can leverage hardware.
As I need this for a fix in the ipv6 stack, which is also included in
this series, I propose to add it sooner than later over Davem's net
tree. This is also the reason why I start slowly with only the one
function I need at this time.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Raw sockets with hdrincl enabled can insert ipv6 extension headers
right into the data stream. In case we need to fragment those packets,
we reparse the options header to find the place where we can insert
the fragment header. If the extension headers exceed the link's MTU we
actually cannot make progress in such a case.
Instead of ending up in broken arithmetic or rounding towards 0 and
entering an endless loop in ip6_fragment, just prevent those cases by
aborting early and signal -EMSGSIZE to user space.
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The idea of the overflow-arith.h header is to collect overflow checking
functions in one central place.
If gcc compiler supports the __builtin_overflow_* builtins we use them
because they might give better performance, otherwise the code falls
back to normal overflow checking functions.
The builtin_overflow functions are supported by gcc-5 and clang. The
matter of supporting clang is to just provide a corresponding
CC_HAVE_BUILTIN_OVERFLOW, because the specific overflow checking builtins
don't differ between gcc and clang.
I just provide overflow_usub function here as I intend this to get merged
into net, more functions will definitely follow as they are needed.
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If alpha is strictly reduced by alpha >> dctcp_shift_g and if alpha is less
than 1 << dctcp_shift_g, then alpha may never reach zero. For example,
given shift_g=4 and alpha=15, alpha >> dctcp_shift_g yields 0 and alpha
remains 15. The effect isn't noticeable in this case below cwnd=137, but
could gradually drive uncongested flows with leftover alpha down to
cwnd=137. A larger dctcp_shift_g would have a greater effect.
This change causes alpha=15 to drop to 0 instead of being decrementing by 1
as it would when alpha=16. However, it requires one less conditional to
implement since it doesn't have to guard against subtracting 1 from 0U. A
decay of 15 is not unreasonable since an equal or greater amount occurs at
alpha >= 240.
Signed-off-by: Andrew G. Shewmaker <agshew@gmail.com>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
The error condition -EAGAIN, which is signaled by throw routes, tells
the rules framework to walk on searching for next matches. If the walk
ends and we stop walking the rules with the result of a throw route we
have to translate the error conditions to -ENETUNREACH.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We don't have fraglist support in TAP_FEATURES. This will lead
software segmentation of gro skb with frag list. Fixes by having
frag list support in TAP_FEATURES.
With this patch single session of netperf receiving were restored from
about 5Gb/s to about 12Gb/s on mlx4.
Fixes a567dd6252 ("macvtap: simplify usage of tap_features")
Cc: Vlad Yasevich <vyasevic@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While transitioning to netdev based vport we broke OVS
feature which allows user to retrieve tunnel packet egress
information for lwtunnel devices. Following patch fixes it
by introducing ndo operation to get the tunnel egress info.
Same ndo operation can be used for lwtunnel devices and compat
ovs-tnl-vport devices. So after adding such device operation
we can remove similar operation from ovs-vport.
Fixes: 614732eaa1 ("openvswitch: Use regular VXLAN net_device device").
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jeff Kirsher says:
====================
Intel Wired LAN Driver Updates 2015-10-22
This series contains fixes to i40e only.
Jesse provides two small fixes for i40e, first fixes counters that were
being displayed incorrectly due to indexing beyond the array of strings
when printing stats. Then fixed the fact that the driver was printing
a message about not being able to assign VMDq because a lack of MSI-X
vectors, when it was not true. It was due to a line missing that
initialized a variable.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The recent fix for the vsock sock_put issue used the wrong
initializer for the transport spin_lock causing an issue when
running with lockdep checking.
Testing: Verified fix on kernel with lockdep enabled.
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Jorgen Hansen <jhansen@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver was printing a message about not being able
to assign VMDq because of a lack of MSI-X vectors.
This was because a line was missing that initialized a variable,
simply a merge error.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>