t.qset_idx can be indirectly controlled by user-space, hence leading to
a potential exploitation of the Spectre variant 1 vulnerability.
This issue was detected with the help of Smatch:
drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c:2286 cxgb_extension_ioctl()
warn: potential spectre issue 'adapter->msix_info'
Fix this by sanitizing t.qset_idx before using it to index
adapter->msix_info
Notice that given that speculation windows are large, the policy is
to kill the speculation on the first load and not worry if it can be
completed with a dependent load/store [1].
[1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2
Cc: stable@vger.kernel.org
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rhashtable_init() currently does not take into account the user-passed
min_size parameter unless param->nelem_hint is set as well. As such,
the default size (number of buckets) will always be HASH_DEFAULT_SIZE
even if the smallest allowed size is larger than that. Remediate this
by unconditionally calling into rounded_hashtable_size() and handling
things accordingly.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ursula Braun says:
====================
net/smc: fixes 2018-07-18
here are small fixes for SMC: The first patch speeds up unidirectional
traffic, the second patch increases security, and the third patch
fixes a problem for fallback cases.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
During clc handshake the receive timeout is set to CLC_WAIT_TIME.
Remember and reset the original timeout value after the receive calls,
and remove a duplicate assignment of CLC_WAIT_TIME.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For security reasons the return code of get_user() should always be
checked.
Fixes: 01d2f7e2cd ("net/smc: sockopts TCP_NODELAY and TCP_CORK")
Reported-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The SMC protocol requires to send a separate consumer cursor update,
if it cannot be piggybacked to updates of the producer cursor.
Currently the decision to send a separate consumer cursor update
just considers the amount of data already received by the socket
program. It does not consider the amount of data already arrived, but
not yet consumed by the receiver. Basing the decision on the
difference between already confirmed and already arrived data
(instead of difference between already confirmed and already consumed
data), may lead to a somewhat earlier consumer cursor update send in
fast unidirectional traffic scenarios, and thus to better throughput.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Suggested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
syzbot is reporting stalls at nfc_llcp_send_ui_frame() [1]. This is
because nfc_llcp_send_ui_frame() is retrying the loop without any delay
when nonblocking nfc_alloc_send_skb() returned NULL.
Since there is no need to use MSG_DONTWAIT if we retry until
sock_alloc_send_pskb() succeeds, let's use blocking call.
Also, in case an unexpected error occurred, let's break the loop
if blocking nfc_alloc_send_skb() failed.
[1] https://syzkaller.appspot.com/bug?id=4a131cc571c3733e0eff6bc673f4e36ae48f19c6
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: syzbot <syzbot+d29d18215e477cfbfbdd@syzkaller.appspotmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
My randconfig builds came across an old missing dependency for ILA:
ERROR: "dst_cache_set_ip6" [net/ipv6/ila/ila.ko] undefined!
ERROR: "dst_cache_get" [net/ipv6/ila/ila.ko] undefined!
ERROR: "dst_cache_init" [net/ipv6/ila/ila.ko] undefined!
ERROR: "dst_cache_destroy" [net/ipv6/ila/ila.ko] undefined!
We almost never run into this by accident because randconfig builds
end up selecting DST_CACHE from some other tunnel protocol, and this
one appears to be the only one missing the explicit 'select'.
>From all I can tell, this problem first appeared in linux-4.9
when dst_cache support got added to ILA.
Fixes: 79ff2fc31e ("ila: Cache a route to translated address")
Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
This driver can spam the kernel log with multiple messages of:
net eth0: eth0: allmulti set
Usually 4 or 8 at a time (probably because of using ConnMan).
This message doesn't seem useful, so let's demote it from dev_info()
to dev_dbg().
Signed-off-by: David Lechner <david@lechnology.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
octeon_mgmt driver doesn't drop RX frames that are 1-4 bytes bigger than
MTU set for the corresponding interface. The problem is in the
AGL_GMX_RX0/1_FRM_MAX register setting, which should not account for VLAN
tagging.
According to Octeon HW manual:
"For tagged frames, MAX increases by four bytes for each VLAN found up to a
maximum of two VLANs, or MAX + 8 bytes."
OCTEON_FRAME_HEADER_LEN "define" is fine for ring buffer management, but
should not be used for AGL_GMX_RX0/1_FRM_MAX.
The problem could be easily reproduced using "ping" command. If affected
system has default MTU 1500, other host (having MTU >= 1504) can
successfully "ping" the affected system with payload size 1473-1476,
resulting in IP packets of size 1501-1504 accepted by the mgmt driver.
Fixed system still accepts IP packets of 1500 bytes even with VLAN tagging,
because the limits are lifted in HW as expected, for every VLAN tag.
Signed-off-by: Alexander Sverdlin <alexander.sverdlin@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SMC ioctl processing requires the sock lock to work properly in
all thinkable scenarios.
Problem has been found with RaceFuzzer and fixes:
KASAN: null-ptr-deref Read in smc_ioctl
Reported-by: Byoungyoung Lee <lifeasageek@gmail.com>
Reported-by: syzbot+35b2c5aa76fd398b9fd4@syzkaller.appspotmail.com
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Siva Reddy Kallam says:
====================
tg3: Update copyright and fix for tx timeout with 5762
First patch:
Update copyright
Second patch:
Add higher cpu clock for 5762
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch has fix for TX timeout while running bi-directional
traffic with 100 Mbps using 5762.
Signed-off-by: Sanjeev Bansal <sanjeevb.bansal@broadcom.com>
Signed-off-by: Siva Reddy Kallam <siva.kallam@broadcom.com>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Siva Reddy Kallam <siva.kallam@broadcom.com>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Testing has uncovered a failure case that is not handled properly. In the
event that a login fails and we are not able to recover on the spot, we
return 0 from do_reset, preventing any error recovery code from being
triggered. Additionally, the state is set to "probed" meaning that when we
are able to trigger the error recovery, the driver always comes up in the
probed state. To handle the case properly, we need to return a failure code
here and set the adapter state to the state that we entered the reset in
indicating the state that we would like to come out of the recovery reset
in.
Signed-off-by: John Allen <jallen@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The skb size calculation in lan78xx_tx_bh is in race with the start_xmit,
which could lead to rare kernel oopses. So protect the whole skb walk with
a spin lock. As a benefit we can unlink the skb directly.
This patch was tested on Raspberry Pi 3B+
Link: https://github.com/raspberrypi/linux/issues/2608
Fixes: 55d7de9de6 ("Microchip's LAN7800 family USB 2/3 to 10/100/1000 Ethernet")
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Floris Bos <bos@je-eigen-domein.nl>
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric reported that reverting the patch that fixed and simplified IPv6
multipath routes means reverting back to invalid userspace notifications.
eg.,
$ ip -6 route add 2001:db8:1::/64 nexthop dev eth0 nexthop dev eth1
only generates a single notification:
2001:db8:1::/64 dev eth0 metric 1024 pref medium
While working on a fix for this problem I found another case that is just
broken completely - a multipath route with a gateway followed by device
followed by gateway:
$ ip -6 ro add 2001:db8:103::/64
nexthop via 2001:db8:1::64
nexthop dev dummy2
nexthop via 2001:db8:3::64
In this case the device only route is dropped completely - no notification
to userpsace but no addition to the FIB either:
$ ip -6 ro ls
2001:db8:1::/64 dev dummy1 proto kernel metric 256 pref medium
2001:db8:2::/64 dev dummy2 proto kernel metric 256 pref medium
2001:db8:3::/64 dev dummy3 proto kernel metric 256 pref medium
2001:db8:103::/64 metric 1024
nexthop via 2001:db8:1::64 dev dummy1 weight 1
nexthop via 2001:db8:3::64 dev dummy3 weight 1 pref medium
fe80::/64 dev dummy1 proto kernel metric 256 pref medium
fe80::/64 dev dummy2 proto kernel metric 256 pref medium
fe80::/64 dev dummy3 proto kernel metric 256 pref medium
Really, IPv6 multipath is just FUBAR'ed beyond repair when it comes to
device only routes, so do not allow it all.
This change will break any scripts relying on the mpath api for insert,
but I don't see any other way to handle the permutations. Besides, since
the routes are added to the FIB as standalone (non-multipath) routes the
kernel is not doing what the user requested, so it might as well tell the
user that.
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David Ahern <dsahern@gmail.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Correct previous bad attempt at allowing sockets to come out of TCP
repair without sending window probes. To avoid changing size of
the repair variable in struct tcp_sock, this lets the decision for
sending probes or not to be made when coming out of repair by
introducing two ways to turn it off.
v2:
* Remove erroneous comment; defines now make behavior clear
Fixes: 70b7ff1302 ("tcp: allow user to create repair socket without window probes")
Signed-off-by: Stefan Baranoff <sbaranoff@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Andrei Vagin <avagin@virtuozzo.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a new rx packet arrives, the rx path will decide whether to reuse
the remainder of the page or not according to one of the below conditions:
1. frag_info->frag_stride == PAGE_SIZE / 2
2. frags->page_offset + frag_info->frag_size > PAGE_SIZE;
The first condition is no met for when XDP is set.
For XDP, page_offset is always set to priv->rx_headroom which is
XDP_PACKET_HEADROOM and frag_info->frag_size is around mtu size + some
padding, still the 2nd release condition will hold since
XDP_PACKET_HEADROOM + 1536 < PAGE_SIZE, as a result the page will not
be released and will be _wrongly_ reused for next free rx descriptor.
In XDP there is an assumption to have a page per packet and reuse can
break such assumption and might cause packet data corruptions.
Fix this by adding an extra condition (!priv->rx_headroom) to the 2nd
case to avoid page reuse when XDP is set, since rx_headroom is set to 0
for non XDP setup and set to XDP_PACKET_HEADROOM for XDP setup.
No additional cache line is required for the new condition.
Fixes: 34db548bfb ("mlx4: add page recycling in receive path")
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Suggested-by: Martin KaFai Lau <kafai@fb.com>
CC: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CC [M] drivers/net/ethernet/freescale/fman/fman.o
In file included from ../drivers/net/ethernet/freescale/fman/fman.c:35:
../include/linux/fsl/guts.h: In function 'guts_set_dmacr':
../include/linux/fsl/guts.h:165:2: error: implicit declaration of function 'clrsetbits_be32' [-Werror=implicit-function-declaration]
clrsetbits_be32(&guts->dmacr, 3 << shift, device << shift);
^~~~~~~~~~~~~~~
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Madalin Bucur <madalin.bucur@nxp.com>
Cc: netdev@vger.kernel.org
Cc: linuxppc-dev@lists.ozlabs.org
Signed-off-by: David S. Miller <davem@davemloft.net>
The netvsc device may need to fallback to running in single queue
mode if host side only wants to support single queue.
Recent change for handling mtu broke this in setup logic.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Fixes: 3ffe64f1a6 ("hv_netvsc: split sub-channel setup into async and sync")
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
During a device failover, there may be latency between the loss
of the current backing device and a notification from firmware that
a failover has occurred. This latency can result in a large amount of
error printouts as firmware returns outgoing traffic with a generic
error code. These are not necessarily errors in this case as the
firmware is busy swapping in a new backing adapter and is not ready
to send packets yet. This patch reclassifies those error codes as
warnings with an explanation that a failover may be pending. All
other return codes will be considered errors.
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit adc176c547 ("ipv6 addrconf: Implemented enhanced DAD (RFC7527)")
added enhanced DAD with a nonce length of 6 bytes. However, RFC7527
doesn't specify the length of the nonce, other than being 6 + 8*k bytes,
with integer k >= 0 (RFC3971 5.3.2). The current implementation simply
assumes that the nonce will always be 6 bytes, but others systems are
free to choose different sizes.
If another system sends a nonce of different length but with the same 6
bytes prefix, it shouldn't be considered as the same nonce. Thus, check
that the length of the received nonce is the same as the length we sent.
Ugly scapy test script running on veth0:
def loop():
pkt=sniff(iface="veth0", filter="icmp6", count=1)
pkt = pkt[0]
b = bytearray(pkt[Raw].load)
b[1] += 1
b += b'\xde\xad\xbe\xef\xde\xad\xbe\xef'
pkt[Raw].load = bytes(b)
pkt[IPv6].plen += 8
# fixup checksum after modifying the payload
pkt[IPv6].payload.cksum -= 0x3b44
if pkt[IPv6].payload.cksum < 0:
pkt[IPv6].payload.cksum += 0xffff
sendp(pkt, iface="veth0")
This should result in DAD failure for any address added to veth0's peer,
but is currently ignored.
Fixes: adc176c547 ("ipv6 addrconf: Implemented enhanced DAD (RFC7527)")
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch remove the following documentation warning
drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:103: warning: Excess function parameter 'priv' description in 'stmmac_axi_setup'
It was introduced in commit afea03656a ("stmmac: rework DMA bus setting and introduce new platform AXI structure")
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A KASAN:use-after-free bug was found related to ip6-erspan
while running selftests/net/ip6_gre_headroom.sh
It happens because of following sequence:
- ipv6hdr pointer is obtained from skb
- skb_cow_head() is called, skb->head memory is reallocated
- old data is accessed using ipv6hdr pointer
skb_cow_head() call was added in e41c7c68ea ("ip6erspan: make sure
enough headroom at xmit."), but looking at the history there was a
chance of similar bug because gre_handle_offloads() and pskb_trim()
can also reallocate skb->head memory. Fixes tag points to commit
which introduced possibility of this bug.
This patch moves ipv6hdr pointer assignment after skb_cow_head() call.
Fixes: 5a963eb61b ("ip6_gre: Add ERSPAN native tunnel support")
Signed-off-by: Prashant Bhole <bhole_prashant_q7@lab.ntt.co.jp>
Reviewed-by: Greg Rose <gvrose8192@gmail.com>
Acked-by: William Tu <u9012063@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On XDP_TX we need to free up the frame only when tun_xdp_tx() returns a
negative value. A positive value indicates that the packet is
successfully enqueued to the ptr_ring, so freeing the page causes
use-after-free.
Fixes: 735fc4054b ("xdp: change ndo_xdp_xmit API to support bulking")
Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Acked-by: Jason Wang <jasowang@redhat.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a spelling typo in bonding.txt
Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the zerocopy sendmsg() path, there are error checks to revert
the zerocopy if we get any error code. syzkaller has discovered
that tls_push_record can return -ECONNRESET, which is fatal, and
happens after the point at which it is safe to revert the iter,
as we've already passed the memory to do_tcp_sendpages.
Previously this code could return -ENOMEM and we would want to
revert the iter, but AFAIK this no longer returns ENOMEM after
a447da7d00 ("tls: fix waitall behavior in tls_sw_recvmsg"),
so we fail for all error codes.
Reported-by: syzbot+c226690f7b3126c5ee04@syzkaller.appspotmail.com
Reported-by: syzbot+709f2810a6a05f11d4d3@syzkaller.appspotmail.com
Signed-off-by: Dave Watson <davejwatson@fb.com>
Fixes: 3c4d755915 ("tls: kernel TLS support")
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Constantine Shulyupin <const@MakeLinux.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
My recent fix for dns_resolver_preparse() printing very long strings was
incomplete, as shown by syzbot which still managed to hit the
WARN_ONCE() in set_precision() by adding a crafted "dns_resolver" key:
precision 50001 too large
WARNING: CPU: 7 PID: 864 at lib/vsprintf.c:2164 vsnprintf+0x48a/0x5a0
The bug this time isn't just a printing bug, but also a logical error
when multiple options ("#"-separated strings) are given in the key
payload. Specifically, when separating an option string into name and
value, if there is no value then the name is incorrectly considered to
end at the end of the key payload, rather than the end of the current
option. This bypasses validation of the option length, and also means
that specifying multiple options is broken -- which presumably has gone
unnoticed as there is currently only one valid option anyway.
A similar problem also applied to option values, as the kstrtoul() when
parsing the "dnserror" option will read past the end of the current
option and into the next option.
Fix these bugs by correctly computing the length of the option name and
by copying the option value, null-terminated, into a temporary buffer.
Reproducer for the WARN_ONCE() that syzbot hit:
perl -e 'print "#A#", "\0" x 50000' | keyctl padd dns_resolver desc @s
Reproducer for "dnserror" option being parsed incorrectly (expected
behavior is to fail when seeing the unknown option "foo", actual
behavior was to read the dnserror value as "1#foo" and fail there):
perl -e 'print "#dnserror=1#foo\0"' | keyctl padd dns_resolver desc @s
Reported-by: syzbot <syzkaller@googlegroups.com>
Fixes: 4a2d789267 ("DNS: If the DNS server returns an error, allow that to be cached [ver #2]")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hangbin Liu says:
====================
multicast: init as INCLUDE when join SSM INCLUDE group
Based on RFC3376 5.1 and RFC3810 6.1, we should init as INCLUDE when join SSM
INCLUDE group. In my first version I only clear the group change record. But
this is not enough as when a new group join, it will init as EXCLUDE and
trigger an filter mode change in ip/ip6_mc_add_src(), which will clear all
source addresses' sf_crcount. This will prevent early joined address sending
state change records if multi source addresses joined at the same time.
In this v2 patchset, I fixed it by directly initializing the mode to INCLUDE
for SSM JOIN_SOURCE_GROUP. I also split the original patch into two separated
patches for IPv4 and IPv6.
Test: test by myself and customer.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This an IPv6 version patch of "ipv4/igmp: init group mode as INCLUDE when
join source group". From RFC3810, part 6.1:
If no per-interface state existed for that
multicast address before the change (i.e., the change consisted of
creating a new per-interface record), or if no state exists after the
change (i.e., the change consisted of deleting a per-interface
record), then the "non-existent" state is considered to have an
INCLUDE filter mode and an empty source list.
Which means a new multicast group should start with state IN(). Currently,
for MLDv2 SSM JOIN_SOURCE_GROUP mode, we first call ipv6_sock_mc_join(),
then ip6_mc_source(), which will trigger a TO_IN() message instead of
ALLOW().
The issue was exposed by commit a052517a8f ("net/multicast: should not
send source list records when have filter mode change"). Before this change,
we sent both ALLOW(A) and TO_IN(A). Now, we only send TO_IN(A).
Fix it by adding a new parameter to init group mode. Also add some wrapper
functions to avoid changing too much code.
v1 -> v2:
In the first version I only cleared the group change record. But this is not
enough. Because when a new group join, it will init as EXCLUDE and trigger
a filter mode change in ip/ip6_mc_add_src(), which will clear all source
addresses sf_crcount. This will prevent early joined address sending state
change records if multi source addressed joined at the same time.
In v2 patch, I fixed it by directly initializing the mode to INCLUDE for SSM
JOIN_SOURCE_GROUP. I also split the original patch into two separated patches
for IPv4 and IPv6.
There is also a difference between v4 and v6 version. For IPv6, when the
interface goes down and up, we will send correct state change record with
unspecified IPv6 address (::) with function ipv6_mc_up(). But after DAD is
completed, we resend the change record TO_IN() in mld_send_initial_cr().
Fix it by sending ALLOW() for INCLUDE mode in mld_send_initial_cr().
Fixes: a052517a8f ("net/multicast: should not send source list records when have filter mode change")
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on RFC3376 5.1
If no interface
state existed for that multicast address before the change (i.e., the
change consisted of creating a new per-interface record), or if no
state exists after the change (i.e., the change consisted of deleting
a per-interface record), then the "non-existent" state is considered
to have a filter mode of INCLUDE and an empty source list.
Which means a new multicast group should start with state IN().
Function ip_mc_join_group() works correctly for IGMP ASM(Any-Source Multicast)
mode. It adds a group with state EX() and inits crcount to mc_qrv,
so the kernel will send a TO_EX() report message after adding group.
But for IGMPv3 SSM(Source-specific multicast) JOIN_SOURCE_GROUP mode, we
split the group joining into two steps. First we join the group like ASM,
i.e. via ip_mc_join_group(). So the state changes from IN() to EX().
Then we add the source-specific address with INCLUDE mode. So the state
changes from EX() to IN(A).
Before the first step sends a group change record, we finished the second
step. So we will only send the second change record. i.e. TO_IN(A).
Regarding the RFC stands, we should actually send an ALLOW(A) message for
SSM JOIN_SOURCE_GROUP as the state should mimic the 'IN() to IN(A)'
transition.
The issue was exposed by commit a052517a8f ("net/multicast: should not
send source list records when have filter mode change"). Before this change,
we used to send both ALLOW(A) and TO_IN(A). After this change we only send
TO_IN(A).
Fix it by adding a new parameter to init group mode. Also add new wrapper
functions so we don't need to change too much code.
v1 -> v2:
In my first version I only cleared the group change record. But this is not
enough. Because when a new group join, it will init as EXCLUDE and trigger
an filter mode change in ip/ip6_mc_add_src(), which will clear all source
addresses' sf_crcount. This will prevent early joined address sending state
change records if multi source addressed joined at the same time.
In v2 patch, I fixed it by directly initializing the mode to INCLUDE for SSM
JOIN_SOURCE_GROUP. I also split the original patch into two separated patches
for IPv4 and IPv6.
Fixes: a052517a8f ("net/multicast: should not send source list records when have filter mode change")
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Yuchung Cheng says:
====================
fix DCTCP delayed ACK
This patch series addresses the issue that sometimes DCTCP
fail to acknowledge the latest sequence and result in sender timeout
if inflight is small.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
After fixing the way DCTCP tracking delayed ACKs, the delayed-ACK
related callbacks are no longer needed
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previously, when a data segment was sent an ACK was piggybacked
on the data segment without generating a CA_EVENT_NON_DELAYED_ACK
event to notify congestion control modules. So the DCTCP
ca->delayed_ack_reserved flag could incorrectly stay set when
in fact there were no delayed ACKs being reserved. This could result
in sending a special ECN notification ACK that carries an older
ACK sequence, when in fact there was no need for such an ACK.
DCTCP keeps track of the delayed ACK status with its own separate
state ca->delayed_ack_reserved. Previously it may accidentally cancel
the delayed ACK without updating this field upon sending a special
ACK that carries a older ACK sequence. This inconsistency would
lead to DCTCP receiver never acknowledging the latest data until the
sender times out and retry in some cases.
Packetdrill script (provided by Larry Brakmo)
0.000 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
0.000 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
0.000 setsockopt(3, SOL_TCP, TCP_CONGESTION, "dctcp", 5) = 0
0.000 bind(3, ..., ...) = 0
0.000 listen(3, 1) = 0
0.100 < [ect0] SEW 0:0(0) win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 7>
0.100 > SE. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
0.110 < [ect0] . 1:1(0) ack 1 win 257
0.200 accept(3, ..., ...) = 4
0.200 < [ect0] . 1:1001(1000) ack 1 win 257
0.200 > [ect01] . 1:1(0) ack 1001
0.200 write(4, ..., 1) = 1
0.200 > [ect01] P. 1:2(1) ack 1001
0.200 < [ect0] . 1001:2001(1000) ack 2 win 257
0.200 write(4, ..., 1) = 1
0.200 > [ect01] P. 2:3(1) ack 2001
0.200 < [ect0] . 2001:3001(1000) ack 3 win 257
0.200 < [ect0] . 3001:4001(1000) ack 3 win 257
0.200 > [ect01] . 3:3(0) ack 4001
0.210 < [ce] P. 4001:4501(500) ack 3 win 257
+0.001 read(4, ..., 4500) = 4500
+0 write(4, ..., 1) = 1
+0 > [ect01] PE. 3:4(1) ack 4501
+0.010 < [ect0] W. 4501:5501(1000) ack 4 win 257
// Previously the ACK sequence below would be 4501, causing a long RTO
+0.040~+0.045 > [ect01] . 4:4(0) ack 5501 // delayed ack
+0.311 < [ect0] . 5501:6501(1000) ack 4 win 257 // More data
+0 > [ect01] . 4:4(0) ack 6501 // now acks everything
+0.500 < F. 9501:9501(0) ack 4 win 257
Reported-by: Larry Brakmo <brakmo@fb.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Lawrence Brakmo <brakmo@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We accidentally left out the error handling for kstrtoul().
Fixes: a520030e32 ("qlcnic: Implement flash sysfs callback for 83xx adapter")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Borkmann says:
====================
pull-request: bpf 2018-07-13
The following pull-request contains BPF updates for your *net* tree.
The main changes are:
1) Fix AF_XDP TX error reporting before final kernel release such that it
becomes consistent between copy mode and zero-copy, from Magnus.
2) Fix three different syzkaller reported issues: oob due to ld_abs
rewrite with too large offset, another oob in l3 based skb test run
and a bug leaving mangled prog in subprog JITing error path, from Daniel.
3) Fix BTF handling for bitfield extraction on big endian, from Okash.
4) Fix a missing linux/errno.h include in cgroup/BPF found by kbuild bot,
from Roman.
5) Fix xdp2skb_meta.sh sample by using just command names instead of
absolute paths for tc and ip and allow them to be redefined, from Taeung.
6) Fix availability probing for BPF seg6 helpers before final kernel ships
so they can be detected at prog load time, from Mathieu.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 8b7008620b ("net: Don't copy pfmemalloc flag in
__copy_skb_header()") introduced a different handling for the
pfmemalloc flag in copy and clone paths.
In __skb_clone(), now, the flag is set only if it was set in the
original skb, but not cleared if it wasn't. This is wrong and
might lead to socket buffers being flagged with pfmemalloc even
if the skb data wasn't allocated from pfmemalloc reserves. Copy
the flag instead of ORing it.
Reported-by: Sabrina Dubroca <sd@queasysnail.net>
Fixes: 8b7008620b ("net: Don't copy pfmemalloc flag in __copy_skb_header()")
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Tested-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Magnus Karlsson says:
====================
This patch set adjusts the AF_XDP TX error reporting so that it becomes
consistent between copy mode and zero-copy. First some background:
Copy-mode for TX uses the SKB path in which the action of sending the
packet is performed from process context using the sendmsg
syscall. Completions are usually done asynchronously from NAPI mode by
using a TX interrupt. In this mode, send errors can be returned back
through the syscall.
In zero-copy mode both the sending of the packet and the completions
are done asynchronously from NAPI mode for performance reasons. In
this mode, the sendmsg syscall only makes sure that the TX NAPI loop
will be run that performs both the actions of sending and
completing. In this mode it is therefore not possible to return errors
through the sendmsg syscall as the sending is done from the NAPI
loop. Note that it is possible to implement a synchronous send with
our API, but in our benchmarks that made the TX performance drop by
nearly half due to synchronization requirements and cache line
bouncing. But for some netdevs this might be preferable so let us
leave it up to the implementation to decide.
The problem is that the current code base returns some errors in
copy-mode that are not possible to return in zero-copy mode. This
patch set aligns them so that the two modes always return the same
error code. We achieve this by removing some of the errors returned by
sendmsg in copy-mode (and in one case adding an error message for
zero-copy mode) and offering alternative error detection methods that
are consistent between the two modes.
The structure of the patch set is as follows:
Patch 1: removes the ENXIO return code from copy-mode when someone has
forcefully changed the number of queues on the device so that the
queue bound to the socket is no longer available. Just silently stop
sending anything as in zero-copy mode.
Patch 2: stop returning EAGAIN in copy mode when the completion queue
is full as zero-copy does not do this. Instead this situation can be
detected by comparing the head and tail pointers of the completion
queue in both modes. In any case, EAGAIN was not the correct error code
here since no amount of calling sendmsg will solve the problem. Only
consuming one or more messages on the completion queue will fix this.
Patch 3: Always return ENOBUFS from sendmsg if there is no TX queue
configured. This was not the case for zero-copy mode.
Patch 4: stop returning EMSGSIZE when the size of the packet is larger
than the MTU. Just send it to the device so that it will drop it as in
zero-copy mode.
Note that copy-mode can still return EAGAIN in certain circumstances,
but as these conditions cannot occur in zero-copy mode it is fine for
copy-mode to return them.
====================
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch stops returning EMSGSIZE from sendmsg in copy mode when the
size of the packet is larger than the MTU. Just send it to the device
so that it will drop it as in zero-copy mode. This makes the error
reporting consistent between copy mode and zero-copy mode.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch makes sure ENOBUFS is always returned from sendmsg if there
is no TX queue configured. This was not the case for zero-copy
mode. With this patch this error reporting is consistent between copy
mode and zero-copy mode.
Fixes: ac98d8aab6 ("xsk: wire upp Tx zero-copy functions")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch stops returning EAGAIN in TX copy mode when the completion
queue is full as zero-copy does not do this. Instead this situation
can be detected by comparing the head and tail pointers of the
completion queue in both modes. In any case, EAGAIN was not the
correct error code here since no amount of calling sendmsg will solve
the problem. Only consuming one or more messages on the completion
queue will fix this.
With this patch, the error reporting becomes consistent between copy
mode and zero-copy mode.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
This patch removes the ENXIO return code from TX copy-mode when
someone has forcefully changed the number of queues on the device so
that the queue bound to the socket is no longer available. Just
silently stop sending anything as in zero-copy mode so the error
reporting gets consistent between the two modes.
Fixes: 35fcde7f8d ("xsk: support for Tx")
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The udpgso benchmark compares various configurations of UDP and TCP.
Including one that is not upstream, udp zerocopy. This is a leftover
from the earlier RFC patchset.
The test is part of kselftests and run in continuous spinners. Remove
the failing case to make the test start passing.
Fixes: 3a687bef14 ("selftests: udp gso benchmark")
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If variable length link layer headers result in a packet shorter
than dev->hard_header_len, reset the network header offset. Else
skb->mac_len may exceed skb->len after skb_mac_reset_len.
packet_sendmsg_spkt already has similar logic.
Fixes: b84bbaf7a6 ("packet: in packet_snd start writing at link layer allocation")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When pulling the NSH header in nsh_gso_segment, set the mac length
based on the encapsulated packet type.
skb_reset_mac_len computes an offset to the network header, which
here still points to the outer packet:
> skb_reset_network_header(skb);
> [...]
> __skb_pull(skb, nsh_len);
> skb_reset_mac_header(skb); // now mac hdr starts nsh_len == 8B after net hdr
> skb_reset_mac_len(skb); // mac len = net hdr - mac hdr == (u16) -8 == 65528
> [..]
> skb_mac_gso_segment(skb, ..)
Link: http://lkml.kernel.org/r/CAF=yD-KeAcTSOn4AxirAxL8m7QAS8GBBe1w09eziYwvPbbUeYA@mail.gmail.com
Reported-by: syzbot+7b9ed9872dab8c32305d@syzkaller.appspotmail.com
Fixes: c411ed8545 ("nsh: add GSO support")
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The pfmemalloc flag indicates that the skb was allocated from
the PFMEMALLOC reserves, and the flag is currently copied on skb
copy and clone.
However, an skb copied from an skb flagged with pfmemalloc
wasn't necessarily allocated from PFMEMALLOC reserves, and on
the other hand an skb allocated that way might be copied from an
skb that wasn't.
So we should not copy the flag on skb copy, and rather decide
whether to allow an skb to be associated with sockets unrelated
to page reclaim depending only on how it was allocated.
Move the pfmemalloc flag before headers_start[0] using an
existing 1-bit hole, so that __copy_skb_header() doesn't copy
it.
When cloning, we'll now take care of this flag explicitly,
contravening to the warning comment of __skb_clone().
While at it, restore the newline usage introduced by commit
b193722731 ("net: reorganize sk_buff for faster
__copy_skb_header()") to visually separate bytes used in
bitfields after headers_start[0], that was gone after commit
a9e419dc7b ("netfilter: merge ctinfo into nfct pointer storage
area"), and describe the pfmemalloc flag in the kernel-doc
structure comment.
This doesn't change the size of sk_buff or cacheline boundaries,
but consolidates the 15 bits hole before tc_index into a 2 bytes
hole before csum, that could now be filled more easily.
Reported-by: Patrick Talbert <ptalbert@redhat.com>
Fixes: c93bdd0e03 ("netvm: allow skb allocation to use PFMEMALLOC reserves")
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
