Currently vlan device inherits unicast filtering flag from underlying
device. If underlying device doesn't support unicast filter, this will
put vlan device into promiscuous mode when it's stacked.
Tun on IFF_UNICAST_FLT on the vlan device in any case so that it does
not go into promiscuous mode needlessly. If underlying device does not
support unicast filtering, that device will enter promiscuous mode.
Signed-off-by: Zhang Shengju <zhangshengju@cmss.chinamobile.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Alexei Starovoitov says:
====================
bpf_get_stackid() and stack_trace map
This patch set introduces new map type to store stack traces and
corresponding bpf_get_stackid() helper.
BPF programs already can walk the stack via unrolled loop
of bpf_probe_read()s which is ok for simple analysis, but it's
not efficient and limited to <30 frames after that the programs
don't fit into MAX_BPF_STACK. With bpf_get_stackid() helper
the programs can collect up to PERF_MAX_STACK_DEPTH both
user and kernel frames.
Using stack traces as a key in a map turned out to be very useful
for generating flame graphs, off-cpu graphs, waker and chain graphs.
Patch 3 is a simplified version of 'offwaketime' tool which is
described in detail here:
http://brendangregg.com/blog/2016-02-01/linux-wakeup-offwake-profiling.html
Earlier version of this patch were using save_stack_trace() helper,
but 'unreliable' frames add to much noise and two equiavlent
stack traces produce different 'stackid's.
Using lockdep style of storing frames with MAX_STACK_TRACE_ENTRIES is
great for lockdep, but not acceptable for bpf, since the stack_trace
map needs to be freed when user Ctrl-C the tool.
The ftrace style with per_cpu(struct ftrace_stack) is great, but it's
tightly coupled with ftrace ring buffer and has the same 'unreliable'
noise. perf_event's perf_callchain() mechanism is also very efficient
and it only needed minor generalization which is done in patch 1
to be used by bpf stack_trace maps.
Peter, please take a look at patch 1.
If you're ok with it, I'd like to take the whole set via net-next.
Patch 1 - generalization of perf_callchain()
Patch 2 - stack_trace map done as lock-less hashtable without link list
to avoid spinlock on insertion which is critical path when
bpf_get_stackid() helper is called for every task switch event
Patch 3 - offwaketime example
After the patch the 'perf report' for artificial 'sched_bench'
benchmark that doing pthread_cond_wait/signal and 'offwaketime'
example is running in the background:
16.35% swapper [kernel.vmlinux] [k] intel_idle
2.18% sched_bench [kernel.vmlinux] [k] __switch_to
2.18% sched_bench libpthread-2.12.so [.] pthread_cond_signal@@GLIBC_2.3.2
1.72% sched_bench libpthread-2.12.so [.] pthread_mutex_unlock
1.53% sched_bench [kernel.vmlinux] [k] bpf_get_stackid
1.44% sched_bench [kernel.vmlinux] [k] entry_SYSCALL_64
1.39% sched_bench [kernel.vmlinux] [k] __call_rcu.constprop.73
1.13% sched_bench libpthread-2.12.so [.] pthread_mutex_lock
1.07% sched_bench libpthread-2.12.so [.] pthread_cond_wait@@GLIBC_2.3.2
1.07% sched_bench [kernel.vmlinux] [k] hash_futex
1.05% sched_bench [kernel.vmlinux] [k] do_futex
1.05% sched_bench [kernel.vmlinux] [k] get_futex_key_refs.isra.13
The hotest part of bpf_get_stackid() is inlined jhash2, so we may consider
using some faster hash in the future, but it's good enough for now.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
This is simplified version of Brendan Gregg's offwaketime:
This program shows kernel stack traces and task names that were blocked and
"off-CPU", along with the stack traces and task names for the threads that woke
them, and the total elapsed time from when they blocked to when they were woken
up. The combined stacks, task names, and total time is summarized in kernel
context for efficiency.
Example:
$ sudo ./offwaketime | flamegraph.pl > demo.svg
Open demo.svg in the browser as FlameGraph visualization.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
add new map type to store stack traces and corresponding helper
bpf_get_stackid(ctx, map, flags) - walk user or kernel stack and return id
@ctx: struct pt_regs*
@map: pointer to stack_trace map
@flags: bits 0-7 - numer of stack frames to skip
bit 8 - collect user stack instead of kernel
bit 9 - compare stacks by hash only
bit 10 - if two different stacks hash into the same stackid
discard old
other bits - reserved
Return: >= 0 stackid on success or negative error
stackid is a 32-bit integer handle that can be further combined with
other data (including other stackid) and used as a key into maps.
Userspace will access stackmap using standard lookup/delete syscall commands to
retrieve full stack trace for given stackid.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
. avoid walking the stack when there is no room left in the buffer
. generalize get_perf_callchain() to be called from bpf helper
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace individual implementations with the recently introduced
skb_postpush_rcsum() helper.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Tom Herbert <tom@herbertland.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Kan Liang says:
====================
ethtool per queue parameters support
Modern network interface controllers usually support multiple receive
and transmit queues. Each queue may have its own parameters. For
example, Intel XL710/X710 hardware supports per queue interrupt
moderation. However, current ethtool does not support per queue
parameters option. User has to set parameters for the whole NIC.
This series extends ethtool to support per queue parameters option.
Since the support of per queue parameters vary with different cards,
it is impossible to address all cards in one patch. This series only
supports per queue coalesce options on i40e driver. The framework used
in the patch can be easily extended to other cards and parameters.
The lib bitmap needs to be extended to facilitate exchanging queue bitmaps
between user space and kernel space. Two patches from David's latest V8
patch series are also cited in this series. You may refer to
https://lkml.org/lkml/2016/2/9/919 for more details.
Changes since V6:
- Rebase on commit 76d13b5687. Did minor change in patch 6.
Changes since V5:
- Add test_bitmap.c and bitmap.sh in the series. They are forgot
to be added previously.
- Update the first two patches to David's latest V8 version. The changes
include
- bitmap u32 API returns number of bits copied, unit tests updated
- module_exit in test_bitmap
- Also change the mode of bitmap.sh to 755 according to Ben's suggestion
Changes since V4:
- Modify set/get_per_queue_coalesce function description
- Change the queue number to be u32
- Correct an error of calculating coalesce backup buffer address
- Rename queue_num to n_queues
- Don't log error message in __i40e_get_coalesce
Changes since V3:
- Based on David's lib bitmap.
- ETHTOOL_PERQUEUE should be handled before the containing switch
- Make the rollback code unconditional
- some minor changes according to Ben's feedback
Changes since V2:
- Add queue-specific settings for interrupt moderation in i40e
Changes since V1:
- Checking the sub-command number to determine whether the command
requires CAP_NET_ADMIN
- Refine the struct ethtool_per_queue_op and improve the comments
- Use bitmap functions to parse queue mask
- Improve comments
- Use bitmap functions to parse queue mask
- Improve comments
- Add rollback support
- Correct the way to find the vector for specific queue.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
For i40e driver, each vector has its own ITR register. However, there
are no concept of queue-specific settings in the driver proper. Only
global variable is used to store ITR values. That will cause problems
especially when resetting the vector. The specific ITR values could be
lost.
This patch move rx_itr_setting and tx_itr_setting to i40e_ring to store
specific ITR register for each queue.
i40e_get_coalesce and i40e_set_coalesce are also modified accordingly to
support queue-specific settings. To make it compatible with old ethtool,
if user doesn't specify the queue number, i40e_get_coalesce will return
queue 0's value. While i40e_set_coalesce will apply value to all queues.
Signed-off-by: Kan Liang <kan.liang@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch implements sub command ETHTOOL_SCOALESCE for ioctl
ETHTOOL_PERQUEUE. It introduces an interface set_per_queue_coalesce to
set coalesce of each masked queue to device driver. The wanted coalesce
information are stored in "data" for each masked queue, which can copy
from userspace.
If it fails to set coalesce to device driver, the value which already
set to specific queue will be tried to rollback.
Signed-off-by: Kan Liang <kan.liang@intel.com>
Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch implements sub command ETHTOOL_GCOALESCE for ioctl
ETHTOOL_PERQUEUE. It introduces an interface get_per_queue_coalesce to
get coalesce of each masked queue from device driver. Then the interrupt
coalescing parameters will be copied back to user space one by one.
Signed-off-by: Kan Liang <kan.liang@intel.com>
Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introduce a new ioctl ETHTOOL_PERQUEUE for per queue parameters setting.
The following patches will enable some SUB_COMMANDs for per queue
setting.
Signed-off-by: Kan Liang <kan.liang@intel.com>
Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is mainly testing bitmap construction and conversion to/from u32[]
for now.
Tested:
qemu i386, x86_64, ppc, ppc64 BE and LE, ARM.
Signed-off-by: David Decotigny <decot@googlers.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Aimed at transferring bitmaps to/from user-space in a 32/64-bit agnostic
way.
Tested:
unit tests (next patch) on qemu i386, x86_64, ppc, ppc64 BE and LE,
ARM.
Signed-off-by: David Decotigny <decot@googlers.com>
Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
In ipv4, when the machine receives a ICMP_FRAG_NEEDED message, the
connected UDP socket will get EMSGSIZE message on its next read from the
socket.
However, this is not the case for ipv6.
This fix modifies the udp err handler in Ipv6 for ICMP6_PKT_TOOBIG to
make it similar to ipv4 behavior. That is when the machine gets an
ICMP6_PKT_TOOBIG message, the connected UDP socket will get EMSGSIZE
message on its next read from the socket.
Signed-off-by: Wei Wang <weiwan@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Interrupts registered by RoCE driver are not unregistered when
msix interrupts are disabled during error recovery causing a
crash. Detach the adapter instance from RoCE driver when error
is detected to complete the cleanup. Attach the driver again after
the adapter is recovered from error.
Signed-off-by: Padmanabh Ratnakar <padmanabh.ratnakar@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Nikolay Aleksandrov says:
====================
bridge: mdb: add support for extended attributes
This small set allows to extend the per mdb entry exported attributes,
before this set we had only a structure exported which couldn't be changed
because we would've broken user-space, after this we extend the attribute
that was used for the structure and add per-mdb entry attributes after the
struct has been added (see patch 02 for more details). Note that the reason
we can't simply add an attribute after MDBA_MDB_ENTRY_INFO is that current
users (e.g. iproute2) walk over the attribute list directly without
checking for the attribute type.
Patch 01 is a simple change to reduce one indentation level in order to
avoid over 80 char lines.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently mdb entries are exported directly as a structure inside
MDBA_MDB_ENTRY_INFO attribute, we can't really extend it without
breaking user-space. In order to export new mdb fields, I've converted
the MDBA_MDB_ENTRY_INFO into a nested attribute which starts like before
with struct br_mdb_entry (without header, as it's casted directly in
iproute2) and continues with MDBA_MDB_EATTR_ attributes. This way we
keep compatibility with older users and can export new data.
I've tested this with iproute2, both with and without support for the
added attribute and it works fine.
So basically we again have MDBA_MDB_ENTRY_INFO with struct br_mdb_entry
inside but it may contain also some additional MDBA_MDB_EATTR_ attributes
such as MDBA_MDB_EATTR_TIMER which can be parsed by user-space.
So the new structure is:
[MDBA_MDB] = {
[MDBA_MDB_ENTRY] = {
[MDBA_MDB_ENTRY_INFO]
[MDBA_MDB_ENTRY_INFO] { <- Nested attribute
struct br_mdb_entry <- nla_put_nohdr()
[MDBA_MDB_ENTRY attributes] <- normal netlink attributes
}
}
}
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Switch the port check and skip if it's null, this allows us to reduce one
indentation level.
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
bpf_percpu_hash_update() expects rcu lock to be held and warns if it's not,
which pointed out a missing rcu read lock.
Fixes: 15a07b338 ("bpf: add lookup/update support for per-cpu hash and array maps")
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jeff Kirsher says:
====================
40GbE Intel Wired LAN Driver Updates 2016-02-19
This series contains updates to i40e/i40evf only.
Alex Duyck splits up the descriptor count function from the function that
stops the ring to have access to the descriptor count used for the data
portion of the frame. The rewrites the logic for how we determine if we
can transmit the frame or if it needs to be linearized. Place the checksum
close to TSO since they have a lot in common and it can help to reduce the
decision tree for how to handle the frame as the first check in TSO is to
see if checksumming is offloaded.
Carolyn adds functions to blink leds on devices using 10GBaseT PHY since
MAC registers used in other designs do not work in this device configuration.
Fixes an issue where a previously removed message has returned.
Kevin increases the timeout when checking GLGEN_RSTAT_DEVSTATE bit since
linking with particular PHY types, the amount of time it takes for the
GLGEN_RSTAT_DEVSTATE to be set increases greatly.
Neerav changes the receive queues to not wait to be disabled before DCB
has been reconfigured, like transmit queues.
Anjali adds new register definitions for programming the parser, flow
director and RSS blocks in the hardware.
Shannon adds the new opcodes and structures used for asking the firmware
to update receive control registers that need extra care when being
accessed while under heavy traffic. Integrates the new AdminQ functions
for safely accessing the receive control registers that may be affected
by heavy small packet traffic.
Mitch provides another colorful patch description on letting go of
the stale local VSI pointer when the VF resets.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
Zhang Shengju
David S. Miller
Alexei Starovoitov
Daniel Borkmann
Andrew Lunn
Kan Liang
David Decotigny
sixiao@microsoft.com
Wei Wang
Padmanabh Ratnakar
Sergio Prado
Nikolay Aleksandrov
Sasha Levin