On ingress, a network device such as a switch assigns to packets
priority based on various criteria. Common options include interpreting
PCP and DSCP fields according to user configuration. When a packet
egresses the switch, a reverse process may rewrite PCP and/or DSCP
values according to packet priority.
The following three functions support a) obtaining a DSCP-to-priority
map or vice versa, and b) finding default-priority entries in APP
database.
The DCB subsystem supports for APP entries a very generous M:N mapping
between priorities and protocol identifiers. Understandably,
several (say) DSCP values can map to the same priority. But this
asymmetry holds the other way around as well--one priority can map to
several DSCP values. For this reason, the following functions operate in
terms of bitmaps, with ones in positions that match some APP entry.
- dcb_ieee_getapp_dscp_prio_mask_map() to compute for a given netdevice
a map of DSCP-to-priority-mask, which gives for each DSCP value a
bitmap of priorities related to that DSCP value by APP, along the
lines of dcb_ieee_getapp_mask().
- dcb_ieee_getapp_prio_dscp_mask_map() similarly to compute for a given
netdevice a map from priorities to a bitmap of DSCPs.
- dcb_ieee_getapp_default_prio_mask() which finds all default-priority
rules for a given port in APP database, and returns a mask of
priorities allowed by these default-priority rules.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The function dcb_app_lookup walks the list of specified DCB APP entries,
looking for one that matches a given criteria: ifindex, selector,
protocol ID and optionally also priority. The "don't care" value for
priority is set to 0, because that priority has not been allowed under
CEE regime, which predates the IEEE standardization.
Under IEEE, 0 is a valid priority number. But because dcb_app_lookup
considers zero a wild card, attempts to add an APP entry with priority 0
fail when other entries exist for a given ifindex / selector / PID
triplet.
Fix by changing the wild-card value to -1.
Signed-off-by: Petr Machata <petrm@mellanox.com>
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In this patch, we add dcbnl buffer attribute to allow user
change the NIC's buffer configuration such as priority
to buffer mapping and buffer size of individual buffer.
This attribute combined with pfc attribute allows advanced user to
fine tune the qos setting for specific priority queue. For example,
user can give dedicated buffer for one or more priorities or user
can give large buffer to certain priorities.
The dcb buffer configuration will be controlled by lldptool.
lldptool -T -i eth2 -V BUFFER prio 0,2,5,7,1,2,3,6
maps priorities 0,1,2,3,4,5,6,7 to receive buffer 0,2,5,7,1,2,3,6
lldptool -T -i eth2 -V BUFFER size 87296,87296,0,87296,0,0,0,0
sets receive buffer size for buffer 0,1,2,3,4,5,6,7 respectively
After discussion on mailing list with Jakub, Jiri, Ido and John, we agreed to
choose dcbnl over devlink interface since this feature is intended to set
port attributes which are governed by the netdev instance of that port, where
devlink API is more suitable for global ASIC configurations.
We present an use case scenario where dcbnl buffer attribute configured
by advance user helps reduce the latency of messages of different sizes.
Scenarios description:
On ConnectX-5, we run latency sensitive traffic with
small/medium message sizes ranging from 64B to 256KB and bandwidth sensitive
traffic with large messages sizes 512KB and 1MB. We group small, medium,
and large message sizes to their own pfc enables priorities as follow.
Priorities 1 & 2 (64B, 256B and 1KB)
Priorities 3 & 4 (4KB, 8KB, 16KB, 64KB, 128KB and 256KB)
Priorities 5 & 6 (512KB and 1MB)
By default, ConnectX-5 maps all pfc enabled priorities to a single
lossless fixed buffer size of 50% of total available buffer space. The
other 50% is assigned to lossy buffer. Using dcbnl buffer attribute,
we create three equal size lossless buffers. Each buffer has 25% of total
available buffer space. Thus, the lossy buffer size reduces to 25%. Priority
to lossless buffer mappings are set as follow.
Priorities 1 & 2 on lossless buffer #1
Priorities 3 & 4 on lossless buffer #2
Priorities 5 & 6 on lossless buffer #3
We observe improvements in latency for small and medium message sizes
as follows. Please note that the large message sizes bandwidth performance is
reduced but the total bandwidth remains the same.
256B message size (42 % latency reduction)
4K message size (21% latency reduction)
64K message size (16% latency reduction)
CC: Ido Schimmel <idosch@idosch.org>
CC: Jakub Kicinski <jakub.kicinski@netronome.com>
CC: Jiri Pirko <jiri@resnulli.us>
CC: Or Gerlitz <gerlitz.or@gmail.com>
CC: Parav Pandit <parav@mellanox.com>
CC: Aron Silverton <aron.silverton@oracle.com>
Signed-off-by: Huy Nguyen <huyn@mellanox.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
This change allows us to later indicate to rtnetlink core that certain
doit functions should be called without acquiring rtnl_mutex.
This change should have no effect, we simply replace the last (now
unused) calcit argument with the new flag.
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Found by reviewing the warning about unused policy table.
The code implies that it meant to check for size, but since
it unrolled the loop for attribute validation that is never used.
Instead do explicit check for attribute.
Compile tested only. Needs review by original author.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add netlink_ext_ack arg to rtnl_doit_func. Pass extack arg to nlmsg_parse
for doit functions that call it directly.
This is the first step to using extended error reporting in rtnetlink.
>From here individual subsystems can be updated to set netlink_ext_ack as
needed.
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pass the new extended ACK reporting struct to all of the generic
netlink parsing functions. For now, pass NULL in almost all callers
(except for some in the core.)
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In function dcbnl_cee_fill(), returns the value of variable err on
errors. However, on some error paths (e.g. nla put fails), its value may
be 0. It may be better to explicitly set a negative errno to variable
err before returning.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=188881
Signed-off-by: Pan Bian <bianpan2016@163.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The Kconfig currently controlling compilation of this code is:
net/dcb/Kconfig:config DCB
net/dcb/Kconfig: bool "Data Center Bridging support"
...meaning that it currently is not being built as a module by anyone.
Lets remove the modular code that is essentially orphaned, so that
when reading the driver there is no doubt it is builtin-only.
Since module_init translates to device_initcall in the non-modular
case, the init ordering remains unchanged with this commit. We can
change to one of the other priority initcalls (subsys?) at any later
date, if desired.
We also delete the MODULE_LICENSE tag etc. since all that information
is (or is now) already contained at the top of the file in the comments.
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Anish Bhatt <anish@chelsio.com>
Cc: John Fastabend <john.r.fastabend@intel.com>
Cc: Shani Michaeli <shanim@mellanox.com>
Cc: netdev@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As specified in 802.1Qau spec. Add this optional attribute to the
DCB netlink layer. To allow for application to use the new attribute,
NIC drivers should implement and register the callbacks ieee_getqcn,
ieee_setqcn and ieee_getqcnstats.
The QCN attribute holds a set of parameters for management, and
a set of statistics to provide informative data on Congestion-Control
defined by this spec.
Signed-off-by: Shani Michaeli <shanim@mellanox.com>
Signed-off-by: Shachar Raindel <raindel@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Solves possible lockup issues that can be seen from firmware DCB agents calling
into the DCB app api.
DCB firmware event queues can be tied in with NAPI so that dcb events are
generated in softIRQ context. This can results in calls to dcb_*app()
functions which try to take the dcb_lock.
If the the event triggers while we also have the dcb_lock because lldpad or
some other agent happened to be issuing a get/set command we could see a cpu
lockup.
This code was not originally written with firmware agents in mind, hence
grabbing dcb_lock from softIRQ context was not considered.
Signed-off-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Current explanation of dcb_app->priority is wrong. It says priority is
expected to be a 3-bit unsigned integer which is only true when working with
DCBx-IEEE. Use of dcb_app->priority by DCBx-CEE expects it to be 802.1p user
priority bitmap. Updated accordingly
This affects the cxgb4 driver, but I will post those changes as part of a
larger changeset shortly.
Fixes: 3e29027af4 ("dcbnl: add support for ieee8021Qaz attributes")
Signed-off-by: Anish Bhatt <anish@chelsio.com>
Acked-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
v2: fixed issue with checking return of dcbnl_rtnl_ops->getapp()
Signed-off-by: Anish Bhatt <anish@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is possible by passing a netlink socket to a more privileged
executable and then to fool that executable into writing to the socket
data that happens to be valid netlink message to do something that
privileged executable did not intend to do.
To keep this from happening replace bare capable and ns_capable calls
with netlink_capable, netlink_net_calls and netlink_ns_capable calls.
Which act the same as the previous calls except they verify that the
opener of the socket had the desired permissions as well.
Reported-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The following call chain indicates that dcb_doit() is protected
under rtnl_lock. So if we use __dev_get_by_name() instead of
dev_get_by_name() to find interface handlers in it, this would
help us avoid to change interface reference counter.
rtnetlink_rcv()
rtnl_lock()
netlink_rcv_skb()
dcb_doit()
rtnl_unlock()
Cc: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Several files refer to an old address for the Free Software Foundation
in the file header comment. Resolve by replacing the address with
the URL <http://www.gnu.org/licenses/> so that we do not have to keep
updating the header comments anytime the address changes.
CC: John Fastabend <john.r.fastabend@intel.com>
CC: Alex Duyck <alexander.h.duyck@intel.com>
CC: Marcel Holtmann <marcel@holtmann.org>
CC: Gustavo Padovan <gustavo@padovan.org>
CC: Johan Hedberg <johan.hedberg@gmail.com>
CC: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With decnet converted, we can finally get rid of rta_buf and its
computations around it. It also gets rid of the minimal header
length verification since all message handlers do that explicitly
anyway.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
The dcb netlink interface leaks stack memory in various places:
* perm_addr[] buffer is only filled at max with 12 of the 32 bytes but
copied completely,
* no in-kernel driver fills all fields of an IEEE 802.1Qaz subcommand,
so we're leaking up to 58 bytes for ieee_ets structs, up to 136 bytes
for ieee_pfc structs, etc.,
* the same is true for CEE -- no in-kernel driver fills the whole
struct,
Prevent all of the above stack info leaks by properly initializing the
buffers/structures involved.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow DCB and net namespace to work together. This is useful if you
have containers that are bound to 'phys' interfaces that want to
also manage their DCB attributes.
The net namespace is taken from sock_net(skb->sk) of the netlink skb.
CC: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
- In rtnetlink_rcv_msg convert the capable(CAP_NET_ADMIN) check
to ns_capable(net->user-ns, CAP_NET_ADMIN). Allowing unprivileged
users to make netlink calls to modify their local network
namespace.
- In the rtnetlink doit methods add capable(CAP_NET_ADMIN) so
that calls that are not safe for unprivileged users are still
protected.
Later patches will remove the extra capable calls from methods
that are safe for unprivilged users.
Acked-by: Serge Hallyn <serge.hallyn@canonical.com>
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is a frequent mistake to confuse the netlink port identifier with a
process identifier. Try to reduce this confusion by renaming fields
that hold port identifiers portid instead of pid.
I have carefully avoided changing the structures exported to
userspace to avoid changing the userspace API.
I have successfully built an allyesconfig kernel with this change.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A small regression was introduced in the reply command of
dcbnl_pg_setcfg(). User space apps may be expecting the
DCB_ATTR_PG_CFG attribute to be returned with the patch
below TX or RX variants are returned.
commit 7be994138b
Author: Thomas Graf <tgraf@suug.ch>
Date: Wed Jun 13 02:54:55 2012 +0000
dcbnl: Shorten all command handling functions
This patch reverts this behavior and returns DCB_ATTR_PG_CFG
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
EMSGSIZE - ran out of space while constructing message
EOPNOTSUPP - driver/hardware does not support operation
ENODEV - network device not found
EINVAL - invalid message
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allocating and sending the skb in dcb_doit() allows for much
shorter and cleaner command handling functions.
The huge switch statement is replaced with an array based definition
of the handling function and reply message type.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no need to allocate and send the reply message in each
handling function separately. Instead, the reply skb can be allocated
and sent in dcb_doit() directly.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds code to trigger CEE events when an APP change or setall
command is made from user space. This simplifies user space code
significantly by creating a single interface to listen on that
works with both firmware and userland agents.
And if we end up with multiple agents this keeps every thing in
sync userland agents, firmware agents, and kernel notifier consumers.
For an example agent that listens for these events see:
https://github.com/jrfastab/cgdcbxd
cgdcbxd is a daemon used to monitor DCB netlink events and manage
the net_prio control group sub-system.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Shmulik Ravid <shmulikr@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Although not specified in 8021Qaz spec, it could be useful to enable drivers
whose HW supports setting a rate limit for an ETS TC. This patch adds this
optional attribute to DCB netlink. To use it, drivers should implement and
register the callbacks ieee_setmaxrate and ieee_getmaxrate. The units are 64
bits long and specified in Kbps to enable usage over both slow and very fast
networks.
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These macros contain a hidden goto, and are thus extremely error
prone and make code hard to audit.
Signed-off-by: David S. Miller <davem@davemloft.net>
With calls to modular infrastructure, these files really
needs the full module.h header. Call it out so some of the
cleanups of implicit and unrequired includes elsewhere can be
cleaned up.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Add DCBX mode to event notifiers so listeners can learn
currently enabled mode.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use ifindex instead of ifname in the DCB app ring. This makes for a smaller
data structure and faster comparisons. It also avoids possible issues when
a net device is renamed.
Signed-off-by: Mark Rustad <mark.d.rustad@intel.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We need to release "dcb_lock" which we took on the previous line.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch add an unsolicited notification of the DCBX negotiated
parameters for the CEE flavor of the DCBX protocol. The notification
message is identical to the aggregated CEE get operation and holds all
the pertinent local and peer information. The notification routine is
exported so it can be invoked by drivers supporting an embedded DCBX
stack.
Signed-off-by: Shmulik Ravid <shmulikr@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The following couple of patches add dcbnl an unsolicited notification of
the the DCB configuration for the CEE flavor of the DCBX protocol. This
is useful when the user-mode DCB client is not responsible for
conducting and resolving the DCBX negotiation (either because the DCBX
stack is embedded in the HW or the negotiation is handled by another
agent in the host), but still needs to get the negotiated parameters.
This functionality already exists for the IEEE flavor of the DCBX
protocol and these patches add it to the older CEE flavor.
The first patch extends the CEE attribute GET operation to include not
only the peer information, but also all the pertinent local
configuration (negotiated parameters). The second patch adds and export
a CEE specific notification routine.
Signed-off-by: Shmulik Ravid <shmulikr@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
These sk_buff structs were allocated with nlmsg_new() so they should
be freed with nlmsg_free().
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Missing error checking before nla_parse_nested().
Reported-by: Mark Rustad <mark.d.rustad@intel.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Incorrect return type on dcb_setapp() this routine
returns negative error codes. All call sites of
dcb_setapp() assign the return value to an int already
so no need to update drivers.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With multiple APP entries per selector and protocol drivers
or stacks may want to pick a specific value or stripe traffic
across many priorities. Also if an APP entry in use is
deleted the stack/driver may want to choose from the existing
APP entries.
To facilitate this and avoid having duplicate code to walk
the APP ring provide a routine dcb_ieee_getapp_mask() to
return a u8 bitmask of all priorities set for the specified
selector and protocol. This routine and bitmask is a helper
for DCB kernel users.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that we allow multiple IEEE App entries we need a way
to remove specific entries. To do this add the ieee_dcb_delapp()
routine.
Additionaly drivers may need to remove the APP entry from
their firmware tables. Add dcb ops routine to handle this.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds a setapp routine for IEEE802.1Qaz encoded APP data types.
The IEEE 802.1Qaz spec encodes the priority bits differently and
allows for multiple APP data entries of the same selector and
protocol. Trying to force these to use the same set routines was
becoming tedious. Furthermore, userspace could probably enforce
the correct semantics, but expecting drivers to do this seems
error prone in the firmware case.
For these reasons add ieee_dcb_setapp() that understands the
IEEE 802.1Qaz encoded form.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now that dcbnl is being used in many cases by more
than a single agent it is beneficial to be notified
when some entity either driver or user space has
changed the DCB attributes.
Today applications either end up polling the interface
or relying on a user space database to maintain the DCB
state and post events. Polling is a poor solution for
obvious reasons. And relying on a user space database
has its own downside. Namely it has created strange
boot dependencies requiring the database be populated
before any applications dependent on DCB attributes
starts or the application goes into a polling loop.
Populating the database requires negotiating link
setting with the peer and can take anywhere from less
than a second up to a few seconds depending on the switch
implementation.
Perhaps more importantly if another application or an
embedded agent sets a DCB link attribute the database
has no way of knowing other than polling the kernel.
This prevents applications from responding quickly to
changes in link events which at least in the FCoE case
and probably any other protocols expecting a lossless
link may result in IO errors.
By adding a multicast group for DCB we have clean way
to disseminate kernel DCB link attributes up to user
space. Avoiding the need for user space to maintain
a coherant database and disperse events that potentially
do not reflect the current link state.
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>