* [PATCH net-next 2/2] pfifo_fast: drop unneeded additional lock on dequeue
From: Paolo Abeni @ 2018-05-15 14:24 UTC (permalink / raw)
To: netdev
Cc: David S. Miller, Jamal Hadi Salim, Cong Wang, Jiri Pirko,
John Fastabend, Michael S. Tsirkin
In-Reply-To: <cover.1526392746.git.pabeni@redhat.com>
After the previous patch, for NOLOCK qdiscs, q->seqlock is
always held when the dequeue() is invoked, we can drop
any additional locking to protect such operation.
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
include/linux/skb_array.h | 5 +++++
net/sched/sch_generic.c | 4 ++--
2 files changed, 7 insertions(+), 2 deletions(-)
diff --git a/include/linux/skb_array.h b/include/linux/skb_array.h
index a6b6e8bb3d7b..62d9b0a6329f 100644
--- a/include/linux/skb_array.h
+++ b/include/linux/skb_array.h
@@ -97,6 +97,11 @@ static inline bool skb_array_empty_any(struct skb_array *a)
return ptr_ring_empty_any(&a->ring);
}
+static inline struct sk_buff *__skb_array_consume(struct skb_array *a)
+{
+ return __ptr_ring_consume(&a->ring);
+}
+
static inline struct sk_buff *skb_array_consume(struct skb_array *a)
{
return ptr_ring_consume(&a->ring);
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index a126f16bc30b..760ab1b09f8b 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -656,7 +656,7 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
if (__skb_array_empty(q))
continue;
- skb = skb_array_consume_bh(q);
+ skb = __skb_array_consume(q);
}
if (likely(skb)) {
qdisc_qstats_cpu_backlog_dec(qdisc, skb);
@@ -697,7 +697,7 @@ static void pfifo_fast_reset(struct Qdisc *qdisc)
if (!q->ring.queue)
continue;
- while ((skb = skb_array_consume_bh(q)) != NULL)
+ while ((skb = __skb_array_consume(q)) != NULL)
kfree_skb(skb);
}
--
2.14.3
^ permalink raw reply related
* Re: i.MX6S/DL and QCA8334 switch using DSA driver - CPU port not working
From: Michal Vokáč @ 2018-05-15 14:25 UTC (permalink / raw)
To: Andrew Lunn; +Cc: netdev, Vivien Didelot, Florian Fainelli
In-Reply-To: <20180510142904.GB7698@lunn.ch>
On 10.5.2018 16:29, Andrew Lunn wrote:
> I would probably add code to dump all the qca8k registers. Compare the
> values for your working setup and your non-working setup. Hopefully
> they are not too different, and you can quickly get to the bits which
> matter.
Perfect! Thanks to your suggestion I did that again and much more carefully.
After some tedious comparison I think I finally found the problem.
The RGMII works only if I write 0x7e to the PORT0_STATUS (0x7c) register
from setup. Then I found out that this setup is also described in
Qualcomm QCA8334 Q&A document. When I do that, everything work as expected.
Both PORT0 and PORT6 may be configured as xGMII, xMII and SerDes and their
functions may be exchanged. In all cases the port status register should be
set to 0x7X where X depends on the link speed setup.
Translated into register bits this means:
- clear BIT(12) - disable MAC flow control auto-negotiation (set by default)
- clear BIT(7) - disable MAC Tx flow control in half-duplex (set by default)
- set BIT(6) - use full-duplex
- set BIT(5,4) - enable Rx/Tx flow control
- set BIT(3) - enable Rx MAC - this one is tricky, the bit is described as
R/O in datasheet but it does not work when not set.
- set BIT(2) - enable Tx MAC
- set BIT(1,0) - set speed to 1000Mb
In general the fixed-link subnode is not handled in qca8k driver. That is why
the link speed and flow control is not set properly for the CPU port.
And obviously autonegotiation can not be used here.
I wonder whether there are some users of this driver and what may be their
setup that they are not affected by that?
I would like to have confirmed that I understand it correctly and that
the problem is really in the driver not handling fixed-link.
Then I can try to prepare a patch. It will be a small challenge for me
but I would like to do the work. I will look at the other drivers where
I think this is already implemented but any advice on how this should be
done will be appreciated.
Please confirm or correct my suggestions.
Thanks, Michal
^ permalink raw reply
* 'Adding' a writable proc file under /proc/net/afs/ [was [PATCH 3/3] afs: Implement namespacing]
From: David Howells @ 2018-05-15 14:28 UTC (permalink / raw)
To: Christoph Hellwig
Cc: dhowells, linux-afs, linux-kernel, netdev, linux-fsdevel
In-Reply-To: <20180515135333.GA31296@lst.de>
Christoph Hellwig <hch@lst.de> wrote:
> So the real question here is why afs needs a write callback as the first
> user of /proc/net ever in a day and time were we've deprecated adding
> new proc files. This is a discussion that should be had with linux-fsdevel
> and netdev in Cc.
/proc/fs/afs/cells and /proc/fs/afs/rootcell are not new proc files, nor is
the ability to configure the in-kernel afs filesystem by writing to them new.
The cells file has been there since 2002 and the rootcell file before the
start of the git history in 2005 - it's just that they've been in
/proc/fs/afs/ not /proc/net/afs/.
However, the afs procfiles need to be somewhere under /proc/net/ for the
network namespacing, so I moved the directory over and emplaced a symlink at
/proc/fs/afs.
David
^ permalink raw reply
* RE: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Steve Wise @ 2018-05-15 14:31 UTC (permalink / raw)
To: 'Jason Gunthorpe'
Cc: 'Leon Romanovsky', dsahern, stephen, netdev, linux-rdma
In-Reply-To: <20180515135335.GA5615@ziepe.ca>
> From: Jason Gunthorpe <jgg@ziepe.ca>
> On Tue, May 15, 2018 at 08:18:51AM -0500, Steve Wise wrote:
> >
> > > > On Mon, May 14, 2018 at 05:04:26PM -0500, Steve Wise wrote:
> > > > >
> > > > >
> > > > > On 5/14/2018 3:41 PM, Jason Gunthorpe wrote:
> > > > > > On Mon, May 07, 2018 at 08:53:16AM -0700, Steve Wise wrote:
> > > > > >> This enhancement allows printing rdma device-specific state, if
> > > provided
> > > > > >> by the kernel. This is done in a generic manner, so rdma tool
> > doesn't
> > > > > >> need to know about the details of every type of rdma device.
> > > > > >>
> > > > > >> Driver attributes for a rdma resource are in the form of <key,
> > > > > >> [print_type], value> tuples, where the key is a string and the
> > value can
> > > > > >> be any supported driver attribute. The print_type attribute, if
> > present,
> > > > > >> provides a print format to use vs the standard print format for the
> > > type.
> > > > > >> For example, the default print type for a PROVIDER_S32 value is
> "%d
> > ",
> > > > > >> but "0x%x " if the print_type of PRINT_TYPE_HEX is included inthe
> > > tuple.
> > > > > >>
> > > > > >> Driver resources are only printed when the -dd flag is present.
> > > > > >> If -p is present, then the output is formatted to not exceed 80
> > > columns,
> > > > > >> otherwise it is printed as a single row to be grep/awk friendly.
> > > > > >>
> > > > > >> Example output:
> > > > > >>
> > > > > >> # rdma resource show qp lqpn 1028 -dd -p
> > > > > >> link cxgb4_0/- lqpn 1028 rqpn 0 type RC state RTS rq-psn 0 sq-psn 0
> > > > path-mig-state MIGRATED pid 0 comm [nvme_rdma]
> > > > > >> sqid 1028 flushed 0 memsize 123968 cidx 85 pidx 85 wq_pidx 106
> > > > flush_cidx 85 in_use 0
> > > > > >> size 386 flags 0x0 rqid 1029 memsize 16768 cidx 43 pidx 41
> > wq_pidx
> > > > 171 msn 44 rqt_hwaddr 0x2a8a5d00
> > > > > >> rqt_size 256 in_use 128 size 130 idx 43 wr_id
> > 0xffff881057c03408 idx
> > > > 40 wr_id 0xffff881057c033f0
> > > > > > Hey some of these look like kernel pointers.. That is a no-no.. What
> > > > > > is up there?
> > > > >
> > > > > Nothing is defined as a kernel pointer. But wr_id is often a pointer
> > to
> > > > > the kernel rdma application's context...
> > > > >
> > > > > > The wr_id often contains a pointer, right? So we cannot just pass it
> > > > > > to user space..
> > > > >
> > > > > Hmm. It is useful for debugging kernel rdma applications. Perhaps
> > > > > these attrs can be only be sent up by the kernel if the capabilities
> > > > > allow. But previous review comments of the kernel series, which is
> > now
> > > > > merged, forced me to remove passing the capabilities information to
> > the
> > > > > driver resource fill functions.
> > > > >
> > > > > So what's the right way to do this?
> > > >
> > > > The reviewer asked do not pass to drivers whole CAP_.. bits, because
> > > > they anyway don't need such granularity.
> > > >
> > >
> > > Ok thanks.
> >
> > How's this?
> >
> > diff --git a/include/rdma/restrack.h b/include/rdma/restrack.h
> > index 6379685..2cf9c5c 100644
> > +++ b/include/rdma/restrack.h
> > @@ -66,7 +66,8 @@ struct rdma_restrack_root {
> > * Allows rdma drivers to add their own restrack attributes.
> > */
> > int (*fill_res_entry)(struct sk_buff *msg,
> > - struct rdma_restrack_entry *entry);
> > + struct rdma_restrack_entry *entry,
> > + bool net_admin_capable);
> > };
>
> cap net admin is not high enough privledge to see unhashed kernel
> pointers. CAP_RAW_IO? Or follow what printk does?
>
Do you mean CAP_NET_RAW? Here's the comments for it:
/* Allow use of RAW sockets */
/* Allow use of PACKET sockets */
/* Allow binding to any address for transparent proxying (also via NET_ADMIN) */
Func restricted_pointer() from lib/vsprintf.c uses CAP_SYSLOG. The comment for CAP_SYSLOG:
/* Allow configuring the kernel's syslog (printk behaviour) */
Func kallsyms_show_value() also uses CAP_SYSLOG. And there is a non-exported global kptr_restrict that they use apparently to allow overriding all this for profiling.
Here is NET_ADMIN's comments:
/* Allow interface configuration */
/* Allow administration of IP firewall, masquerading and accounting */
/* Allow setting debug option on sockets */
/* Allow modification of routing tables */
/* Allow setting arbitrary process / process group ownership on
sockets */
/* Allow binding to any address for transparent proxying (also via NET_RAW) */
/* Allow setting TOS (type of service) */
/* Allow setting promiscuous mode */
/* Allow clearing driver statistics */
/* Allow multicasting */
/* Allow read/write of device-specific registers */
/* Allow activation of ATM control sockets */
^ permalink raw reply
* Re: [pull request][for-next 0/6] Mellanox, mlx5 updates 2018-05-07
From: Doug Ledford @ 2018-05-15 14:34 UTC (permalink / raw)
To: David Miller, saeedm; +Cc: netdev, linux-rdma, leonro, jgg
In-Reply-To: <20180510.081836.382051449994824032.davem@davemloft.net>
[-- Attachment #1: Type: text/plain, Size: 621 bytes --]
On Thu, 2018-05-10 at 08:18 -0400, David Miller wrote:
> From: Saeed Mahameed <saeedm@mellanox.com>
> Date: Mon, 7 May 2018 16:52:58 -0700
>
> > This pull request includes misc updates and cleanups for mlx5 core
> > driver for both net and rdma next branches, for more information please
> > see tag log below.
> >
> > Please pull and let me know if there's any problem.
>
> Looks good, pulled into net-next.
>
> Thanks.
Thanks, pulled into rdma-next.
--
Doug Ledford <dledford@redhat.com>
GPG KeyID: B826A3330E572FDD
Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDD
[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
^ permalink raw reply
* Re: [PATCH 04/40] proc: introduce proc_create_seq{, _data}
From: Christoph Hellwig @ 2018-05-15 14:42 UTC (permalink / raw)
To: David Howells
Cc: linux-rtc-u79uwXL29TY76Z2rM5mHXA, Alessandro Zummo,
Alexandre Belloni, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-scsi-u79uwXL29TY76Z2rM5mHXA,
linux-acpi-u79uwXL29TY76Z2rM5mHXA, Greg Kroah-Hartman, Jiri Slaby,
megaraidlinux.pdl-dY08KVG/lbpWk0Htik3J/w,
linux-kernel-u79uwXL29TY76Z2rM5mHXA,
linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
linux-ide-u79uwXL29TY76Z2rM5mHXA,
netfilter-devel-u79uwXL29TY76Z2rM5mHXA, Alexander Viro,
netdev-u79uwXL29TY76Z2rM5mHXA, Andrew Morton,
linux-ext4-u79uwXL29TY76Z2rM5mHXA,
drbd-dev-cunTk1MwBs8qoQakbn7OcQ, Christoph Hellwig,
jfs-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f, Alexey Dobriyan
In-Reply-To: <26540.1525094365-S6HVgzuS8uM4Awkfq6JHfwNdhmdF6hFW@public.gmane.org>
On Mon, Apr 30, 2018 at 02:19:25PM +0100, David Howells wrote:
> Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org> wrote:
>
> > +
> > +struct proc_dir_entry *proc_create_seq_data(const char *name, umode_t mode,
> > + struct proc_dir_entry *parent, const struct seq_operations *ops,
> > + void *data)
> > +{
> > ...
> > +EXPORT_SYMBOL(proc_create_seq_data);
>
> Please add documentation comments to exported functions when you add them.
None of the base functions are document, and we really want people to not
use procfs for new code anyway. But if I get some consensus from the
maintainers and the list I can throw in another patch to document
all proc_create* variants.
^ permalink raw reply
* Re: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Jason Gunthorpe @ 2018-05-15 14:44 UTC (permalink / raw)
To: Steve Wise
Cc: 'Leon Romanovsky', dsahern, stephen, netdev, linux-rdma
In-Reply-To: <012201d3ec59$69242fa0$3b6c8ee0$@opengridcomputing.com>
On Tue, May 15, 2018 at 09:31:27AM -0500, Steve Wise wrote:
> > cap net admin is not high enough privledge to see unhashed kernel
> > pointers. CAP_RAW_IO? Or follow what printk does?
> >
>
> Do you mean CAP_NET_RAW? Here's the comments for it:
Nope..
> Func restricted_pointer() from lib/vsprintf.c uses CAP_SYSLOG. The comment for CAP_SYSLOG:
Yikes, yes, that is probably the required logic here, including the
kptr_restrict = 0 thing
Jason
^ permalink raw reply
* Re: [PATCH 11/40] ipv6/flowlabel: simplify pid namespace lookup
From: Christoph Hellwig @ 2018-05-15 14:56 UTC (permalink / raw)
To: Eric W. Biederman
Cc: linux-rtc-u79uwXL29TY76Z2rM5mHXA, Alessandro Zummo,
Alexandre Belloni, devel-gWbeCf7V1WCQmaza687I9mD2FQJk+8+b,
linux-afs-IAPFreCvJWM7uuMidbF8XUB+6BGkLq7r,
linux-scsi-u79uwXL29TY76Z2rM5mHXA,
linux-ide-u79uwXL29TY76Z2rM5mHXA, Greg Kroah-Hartman,
linux-kernel-u79uwXL29TY76Z2rM5mHXA,
jfs-discussion-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f,
Christoph Hellwig, linux-acpi-u79uwXL29TY76Z2rM5mHXA,
netdev-u79uwXL29TY76Z2rM5mHXA,
netfilter-devel-u79uwXL29TY76Z2rM5mHXA, Alexander Viro,
Jiri Slaby, Andrew Morton, linux-ext4-u79uwXL29TY76Z2rM5mHXA,
Alexey Dobriyan, megaraidlinux.pdl-dY08KVG/lbpWk0Htik3J/w,
drbd-dev-cunTk1MwBs8qoQakbn7OcQ
In-Reply-To: <878t8y46sy.fsf-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org>
On Sat, May 05, 2018 at 07:37:33AM -0500, Eric W. Biederman wrote:
> Christoph Hellwig <hch-jcswGhMUV9g@public.gmane.org> writes:
>
> > The shole seq_file sequence already operates under a single RCU lock pair,
> > so move the pid namespace lookup into it, and stop grabbing a reference
> > and remove all kinds of boilerplate code.
>
> This is wrong.
>
> Move task_active_pid_ns(current) from open to seq_start actually means
> that the results if you pass this proc file between callers the results
> will change. So this breaks file descriptor passing.
>
> Open is a bad place to access current. In the middle of read/write is
> broken.
>
>
> In this particular instance looking up the pid namespace with
> task_active_pid_ns was a personal brain fart. What the code should be
> doing (with an appropriate helper) is:
>
> struct pid_namespace *pid_ns = inode->i_sb->s_fs_info;
>
> Because each mount of proc is bound to a pid namespace. Looking up the
> pid namespace from the super_block is a much better way to go.
What do you have in mind for the helper? For now I've thrown it in
opencoded into my working tree, but I'd be glad to add a helper.
struct pid_namespace *proc_pid_namespace(struct inode *inode)
{
// maybe warn on for s_magic not on procfs??
return inode->i_sb->s_fs_info;
}
?
^ permalink raw reply
* RE: [PATCH v1 iproute2-next 2/3] rdma: print driver resource attributes
From: Steve Wise @ 2018-05-15 15:02 UTC (permalink / raw)
To: 'Jason Gunthorpe'
Cc: 'Leon Romanovsky', dsahern, stephen, netdev, linux-rdma
In-Reply-To: <20180515144420.GB5615@ziepe.ca>
> On Tue, May 15, 2018 at 09:31:27AM -0500, Steve Wise wrote:
> > > cap net admin is not high enough privledge to see unhashed kernel
> > > pointers. CAP_RAW_IO? Or follow what printk does?
> > >
> >
> > Do you mean CAP_NET_RAW? Here's the comments for it:
>
> Nope..
>
> > Func restricted_pointer() from lib/vsprintf.c uses CAP_SYSLOG. The
> comment for CAP_SYSLOG:
>
> Yikes, yes, that is probably the required logic here, including the
> kptr_restrict = 0 thing
>
Let's defer the ktpr_restrict issue for now; I want to finish the initial
work this cycle, and adding that will likely take too much time. I'll use
CAP_SYSLOG and add a FIXME comment. Ok?
Steve.
^ permalink raw reply
* [PATCH net-next 00/10] net/smc: enhancements 2018/05/15
From: Ursula Braun @ 2018-05-15 15:04 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
From: Ursula Braun <ursula.braun@de.ibm.com>
Dave,
here are smc patches for net-next. The first one is a fix for net-next
commit 01d2f7e2cdd3 "net/smc: sockopts TCP_NODELAY and TCP_CORK".
Patch 7 improves Connection Layer Control error handling, patch 10
improves abnormal termination of link groups. The remaining patches
from Karsten improve Link Layer Control code.
Thanks, Ursula
Karsten Graul (9):
net/smc: register new rmbs with the peer
net/smc: remove unnecessary cast
net/smc: simplify test_link function usage
net/smc: move link llc initialization to llc layer
net/smc: use a workqueue to defer llc send
net/smc: handle all error codes from smc_conn_create()
net/smc: set link inactive before calling smc_lgr_free()
net/smc: drop messages when link state is inactive
net/smc: check for pending termination
Ursula Braun (1):
net/smc: no tx work trigger for fallback sockets
net/smc/af_smc.c | 27 +++++---
net/smc/smc_clc.c | 2 +-
net/smc/smc_core.c | 23 ++++---
net/smc/smc_core.h | 6 +-
net/smc/smc_llc.c | 196 +++++++++++++++++++++++++++++++++++++++++------------
net/smc/smc_llc.h | 7 +-
6 files changed, 196 insertions(+), 65 deletions(-)
--
2.16.3
^ permalink raw reply
* [PATCH net-next 01/10] net/smc: no tx work trigger for fallback sockets
From: Ursula Braun @ 2018-05-15 15:04 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
If TCP_NODELAY is set or TCP_CORK is reset, setsockopt triggers the
tx worker. This does not make sense, if the SMC socket switched to
the TCP fallback when the connection is created. This patch adds
the additional check for the fallback case.
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/af_smc.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 17688a02035b..83403be46a4a 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -1353,14 +1353,14 @@ static int smc_setsockopt(struct socket *sock, int level, int optname,
break;
case TCP_NODELAY:
if (sk->sk_state != SMC_INIT && sk->sk_state != SMC_LISTEN) {
- if (val)
+ if (val && !smc->use_fallback)
mod_delayed_work(system_wq, &smc->conn.tx_work,
0);
}
break;
case TCP_CORK:
if (sk->sk_state != SMC_INIT && sk->sk_state != SMC_LISTEN) {
- if (!val)
+ if (!val && !smc->use_fallback)
mod_delayed_work(system_wq, &smc->conn.tx_work,
0);
}
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 02/10] net/smc: register new rmbs with the peer
From: Ursula Braun @ 2018-05-15 15:04 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Register new rmb buffers with the remote peer by exchanging a
confirm_rkey llc message.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/af_smc.c | 20 ++++++++++++++------
net/smc/smc_core.c | 1 +
net/smc/smc_core.h | 2 ++
net/smc/smc_llc.c | 47 +++++++++++++++++++++++++++++++++++++++++++++--
net/smc/smc_llc.h | 2 ++
5 files changed, 64 insertions(+), 8 deletions(-)
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 83403be46a4a..397ba2182453 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -293,14 +293,22 @@ static void smc_copy_sock_settings_to_smc(struct smc_sock *smc)
smc_copy_sock_settings(&smc->sk, smc->clcsock->sk, SK_FLAGS_CLC_TO_SMC);
}
-/* register a new rmb */
-static int smc_reg_rmb(struct smc_link *link, struct smc_buf_desc *rmb_desc)
+/* register a new rmb, optionally send confirm_rkey msg to register with peer */
+static int smc_reg_rmb(struct smc_link *link, struct smc_buf_desc *rmb_desc,
+ bool conf_rkey)
{
/* register memory region for new rmb */
if (smc_wr_reg_send(link, rmb_desc->mr_rx[SMC_SINGLE_LINK])) {
rmb_desc->regerr = 1;
return -EFAULT;
}
+ if (!conf_rkey)
+ return 0;
+ /* exchange confirm_rkey msg with peer */
+ if (smc_llc_do_confirm_rkey(link, rmb_desc)) {
+ rmb_desc->regerr = 1;
+ return -EFAULT;
+ }
return 0;
}
@@ -334,7 +342,7 @@ static int smc_clnt_conf_first_link(struct smc_sock *smc)
smc_wr_remember_qp_attr(link);
- if (smc_reg_rmb(link, smc->conn.rmb_desc))
+ if (smc_reg_rmb(link, smc->conn.rmb_desc, false))
return SMC_CLC_DECL_INTERR;
/* send CONFIRM LINK response over RoCE fabric */
@@ -488,7 +496,7 @@ static int smc_connect_rdma(struct smc_sock *smc)
}
} else {
if (!smc->conn.rmb_desc->reused) {
- if (smc_reg_rmb(link, smc->conn.rmb_desc)) {
+ if (smc_reg_rmb(link, smc->conn.rmb_desc, true)) {
reason_code = SMC_CLC_DECL_INTERR;
goto decline_rdma_unlock;
}
@@ -729,7 +737,7 @@ static int smc_serv_conf_first_link(struct smc_sock *smc)
link = &lgr->lnk[SMC_SINGLE_LINK];
- if (smc_reg_rmb(link, smc->conn.rmb_desc))
+ if (smc_reg_rmb(link, smc->conn.rmb_desc, false))
return SMC_CLC_DECL_INTERR;
/* send CONFIRM LINK request to client over the RoCE fabric */
@@ -866,7 +874,7 @@ static void smc_listen_work(struct work_struct *work)
if (local_contact != SMC_FIRST_CONTACT) {
if (!new_smc->conn.rmb_desc->reused) {
- if (smc_reg_rmb(link, new_smc->conn.rmb_desc)) {
+ if (smc_reg_rmb(link, new_smc->conn.rmb_desc, true)) {
reason_code = SMC_CLC_DECL_INTERR;
goto decline_rdma_unlock;
}
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index 9c74844e2008..29ae077b0ec9 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -210,6 +210,7 @@ static int smc_lgr_create(struct smc_sock *smc,
init_completion(&lnk->llc_confirm_resp);
init_completion(&lnk->llc_add);
init_completion(&lnk->llc_add_resp);
+ init_completion(&lnk->llc_confirm_rkey);
smc->conn.lgr = lgr;
rwlock_init(&lgr->conns_lock);
diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h
index fca8624e5e71..9398e235d74b 100644
--- a/net/smc/smc_core.h
+++ b/net/smc/smc_core.h
@@ -105,6 +105,8 @@ struct smc_link {
struct delayed_work llc_testlink_wrk; /* testlink worker */
struct completion llc_testlink_resp; /* wait for rx of testlink */
int llc_testlink_time; /* testlink interval */
+ struct completion llc_confirm_rkey; /* wait 4 rx of cnf rkey */
+ int llc_confirm_rkey_rc; /* rc from cnf rkey msg */
};
/* For now we just allow one parallel link per link group. The SMC protocol
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 33b4d856f4c6..b118f80b9c41 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -214,6 +214,31 @@ int smc_llc_send_confirm_link(struct smc_link *link, u8 mac[],
return rc;
}
+/* send LLC confirm rkey request */
+static int smc_llc_send_confirm_rkey(struct smc_link *link,
+ struct smc_buf_desc *rmb_desc)
+{
+ struct smc_llc_msg_confirm_rkey *rkeyllc;
+ struct smc_wr_tx_pend_priv *pend;
+ struct smc_wr_buf *wr_buf;
+ int rc;
+
+ rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
+ if (rc)
+ return rc;
+ rkeyllc = (struct smc_llc_msg_confirm_rkey *)wr_buf;
+ memset(rkeyllc, 0, sizeof(*rkeyllc));
+ rkeyllc->hd.common.type = SMC_LLC_CONFIRM_RKEY;
+ rkeyllc->hd.length = sizeof(struct smc_llc_msg_confirm_rkey);
+ rkeyllc->rtoken[0].rmb_key =
+ htonl(rmb_desc->mr_rx[SMC_SINGLE_LINK]->rkey);
+ rkeyllc->rtoken[0].rmb_vaddr = cpu_to_be64(
+ (u64)sg_dma_address(rmb_desc->sgt[SMC_SINGLE_LINK].sgl));
+ /* send llc message */
+ rc = smc_wr_tx_send(link, pend);
+ return rc;
+}
+
/* send ADD LINK request or response */
int smc_llc_send_add_link(struct smc_link *link, u8 mac[],
union ib_gid *gid,
@@ -413,7 +438,9 @@ static void smc_llc_rx_confirm_rkey(struct smc_link *link,
lgr = container_of(link, struct smc_link_group, lnk[SMC_SINGLE_LINK]);
if (llc->hd.flags & SMC_LLC_FLAG_RESP) {
- /* unused as long as we don't send this type of msg */
+ link->llc_confirm_rkey_rc = llc->hd.flags &
+ SMC_LLC_FLAG_RKEY_NEG;
+ complete(&link->llc_confirm_rkey);
} else {
rc = smc_rtoken_add(lgr,
llc->rtoken[0].rmb_vaddr,
@@ -503,7 +530,7 @@ static void smc_llc_rx_handler(struct ib_wc *wc, void *buf)
}
}
-/***************************** worker ****************************************/
+/***************************** worker, utils *********************************/
static void smc_llc_testlink_work(struct work_struct *work)
{
@@ -562,6 +589,22 @@ void smc_llc_link_flush(struct smc_link *link)
cancel_delayed_work_sync(&link->llc_testlink_wrk);
}
+/* register a new rtoken at the remote peer */
+int smc_llc_do_confirm_rkey(struct smc_link *link,
+ struct smc_buf_desc *rmb_desc)
+{
+ int rc;
+
+ reinit_completion(&link->llc_confirm_rkey);
+ smc_llc_send_confirm_rkey(link, rmb_desc);
+ /* receive CONFIRM RKEY response from server over RoCE fabric */
+ rc = wait_for_completion_interruptible_timeout(&link->llc_confirm_rkey,
+ SMC_LLC_WAIT_TIME);
+ if (rc <= 0 || link->llc_confirm_rkey_rc)
+ return -EFAULT;
+ return 0;
+}
+
/***************************** init, exit, misc ******************************/
static struct smc_wr_rx_handler smc_llc_rx_handlers[] = {
diff --git a/net/smc/smc_llc.h b/net/smc/smc_llc.h
index d6e42116485e..728823dd8ae5 100644
--- a/net/smc/smc_llc.h
+++ b/net/smc/smc_llc.h
@@ -47,6 +47,8 @@ int smc_llc_send_test_link(struct smc_link *lnk, u8 user_data[16],
void smc_llc_link_active(struct smc_link *link, int testlink_time);
void smc_llc_link_inactive(struct smc_link *link);
void smc_llc_link_flush(struct smc_link *link);
+int smc_llc_do_confirm_rkey(struct smc_link *link,
+ struct smc_buf_desc *rmb_desc);
int smc_llc_init(void) __init;
#endif /* SMC_LLC_H */
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 06/10] net/smc: use a workqueue to defer llc send
From: Ursula Braun @ 2018-05-15 15:04 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
SMC handles deferred work in tasklets. As tasklets cannot sleep this
can result in rare EBUSY conditions, so defer this work in a work queue.
The high level api functions do not defer work because they can sleep
until the llc send is actually completed.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/smc_core.c | 10 ++--
net/smc/smc_core.h | 1 +
net/smc/smc_llc.c | 132 ++++++++++++++++++++++++++++++++++++++---------------
net/smc/smc_llc.h | 4 +-
4 files changed, 104 insertions(+), 43 deletions(-)
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index ba7dae819ac8..4523bbad8a15 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -194,10 +194,12 @@ static int smc_lgr_create(struct smc_sock *smc,
smc_ib_setup_per_ibdev(smcibdev);
get_random_bytes(rndvec, sizeof(rndvec));
lnk->psn_initial = rndvec[0] + (rndvec[1] << 8) + (rndvec[2] << 16);
- smc_llc_link_init(lnk);
- rc = smc_wr_alloc_link_mem(lnk);
+ rc = smc_llc_link_init(lnk);
if (rc)
goto free_lgr;
+ rc = smc_wr_alloc_link_mem(lnk);
+ if (rc)
+ goto clear_llc_lnk;
rc = smc_ib_create_protection_domain(lnk);
if (rc)
goto free_link_mem;
@@ -221,6 +223,8 @@ static int smc_lgr_create(struct smc_sock *smc,
smc_ib_dealloc_protection_domain(lnk);
free_link_mem:
smc_wr_free_link_mem(lnk);
+clear_llc_lnk:
+ smc_llc_link_clear(lnk);
free_lgr:
kfree(lgr);
out:
@@ -266,6 +270,7 @@ void smc_conn_free(struct smc_connection *conn)
static void smc_link_clear(struct smc_link *lnk)
{
lnk->peer_qpn = 0;
+ smc_llc_link_clear(lnk);
smc_ib_modify_qp_reset(lnk);
smc_wr_free_link(lnk);
smc_ib_destroy_queue_pair(lnk);
@@ -323,7 +328,6 @@ static void smc_lgr_free_bufs(struct smc_link_group *lgr)
/* remove a link group */
void smc_lgr_free(struct smc_link_group *lgr)
{
- smc_llc_link_flush(&lgr->lnk[SMC_SINGLE_LINK]);
smc_lgr_free_bufs(lgr);
smc_link_clear(&lgr->lnk[SMC_SINGLE_LINK]);
kfree(lgr);
diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h
index 9398e235d74b..d1753850684b 100644
--- a/net/smc/smc_core.h
+++ b/net/smc/smc_core.h
@@ -96,6 +96,7 @@ struct smc_link {
u8 link_id; /* unique # within link group */
enum smc_link_state state; /* state of link */
+ struct workqueue_struct *llc_wq; /* single thread work queue */
struct completion llc_confirm; /* wait for rx of conf link */
struct completion llc_confirm_resp; /* wait 4 rx of cnf lnk rsp */
int llc_confirm_rc; /* rc from confirm link msg */
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 5dbeced6cea5..5d7926cc472c 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -239,6 +239,25 @@ static int smc_llc_send_confirm_rkey(struct smc_link *link,
return rc;
}
+/* prepare an add link message */
+static void smc_llc_prep_add_link(struct smc_llc_msg_add_link *addllc,
+ struct smc_link *link, u8 mac[],
+ union ib_gid *gid,
+ enum smc_llc_reqresp reqresp)
+{
+ memset(addllc, 0, sizeof(*addllc));
+ addllc->hd.common.type = SMC_LLC_ADD_LINK;
+ addllc->hd.length = sizeof(struct smc_llc_msg_add_link);
+ if (reqresp == SMC_LLC_RESP) {
+ addllc->hd.flags |= SMC_LLC_FLAG_RESP;
+ /* always reject more links for now */
+ addllc->hd.flags |= SMC_LLC_FLAG_ADD_LNK_REJ;
+ addllc->hd.add_link_rej_rsn = SMC_LLC_REJ_RSN_NO_ALT_PATH;
+ }
+ memcpy(addllc->sender_mac, mac, ETH_ALEN);
+ memcpy(addllc->sender_gid, gid, SMC_GID_SIZE);
+}
+
/* send ADD LINK request or response */
int smc_llc_send_add_link(struct smc_link *link, u8 mac[],
union ib_gid *gid,
@@ -253,22 +272,28 @@ int smc_llc_send_add_link(struct smc_link *link, u8 mac[],
if (rc)
return rc;
addllc = (struct smc_llc_msg_add_link *)wr_buf;
- memset(addllc, 0, sizeof(*addllc));
- addllc->hd.common.type = SMC_LLC_ADD_LINK;
- addllc->hd.length = sizeof(struct smc_llc_msg_add_link);
- if (reqresp == SMC_LLC_RESP) {
- addllc->hd.flags |= SMC_LLC_FLAG_RESP;
- /* always reject more links for now */
- addllc->hd.flags |= SMC_LLC_FLAG_ADD_LNK_REJ;
- addllc->hd.add_link_rej_rsn = SMC_LLC_REJ_RSN_NO_ALT_PATH;
- }
- memcpy(addllc->sender_mac, mac, ETH_ALEN);
- memcpy(addllc->sender_gid, gid, SMC_GID_SIZE);
+ smc_llc_prep_add_link(addllc, link, mac, gid, reqresp);
/* send llc message */
rc = smc_wr_tx_send(link, pend);
return rc;
}
+/* prepare a delete link message */
+static void smc_llc_prep_delete_link(struct smc_llc_msg_del_link *delllc,
+ struct smc_link *link,
+ enum smc_llc_reqresp reqresp)
+{
+ memset(delllc, 0, sizeof(*delllc));
+ delllc->hd.common.type = SMC_LLC_DELETE_LINK;
+ delllc->hd.length = sizeof(struct smc_llc_msg_add_link);
+ if (reqresp == SMC_LLC_RESP)
+ delllc->hd.flags |= SMC_LLC_FLAG_RESP;
+ /* DEL_LINK_ALL because only 1 link supported */
+ delllc->hd.flags |= SMC_LLC_FLAG_DEL_LINK_ALL;
+ delllc->hd.flags |= SMC_LLC_FLAG_DEL_LINK_ORDERLY;
+ delllc->link_num = link->link_id;
+}
+
/* send DELETE LINK request or response */
int smc_llc_send_delete_link(struct smc_link *link,
enum smc_llc_reqresp reqresp)
@@ -282,15 +307,7 @@ int smc_llc_send_delete_link(struct smc_link *link,
if (rc)
return rc;
delllc = (struct smc_llc_msg_del_link *)wr_buf;
- memset(delllc, 0, sizeof(*delllc));
- delllc->hd.common.type = SMC_LLC_DELETE_LINK;
- delllc->hd.length = sizeof(struct smc_llc_msg_add_link);
- if (reqresp == SMC_LLC_RESP)
- delllc->hd.flags |= SMC_LLC_FLAG_RESP;
- /* DEL_LINK_ALL because only 1 link supported */
- delllc->hd.flags |= SMC_LLC_FLAG_DEL_LINK_ALL;
- delllc->hd.flags |= SMC_LLC_FLAG_DEL_LINK_ORDERLY;
- delllc->link_num = link->link_id;
+ smc_llc_prep_delete_link(delllc, link, reqresp);
/* send llc message */
rc = smc_wr_tx_send(link, pend);
return rc;
@@ -317,20 +334,46 @@ static int smc_llc_send_test_link(struct smc_link *link, u8 user_data[16])
return rc;
}
-/* send a prepared message */
-static int smc_llc_send_message(struct smc_link *link, void *llcbuf, int llclen)
+struct smc_llc_send_work {
+ struct work_struct work;
+ struct smc_link *link;
+ int llclen;
+ union smc_llc_msg llcbuf;
+};
+
+/* worker that sends a prepared message */
+static void smc_llc_send_message_work(struct work_struct *work)
{
+ struct smc_llc_send_work *llcwrk = container_of(work,
+ struct smc_llc_send_work, work);
struct smc_wr_tx_pend_priv *pend;
struct smc_wr_buf *wr_buf;
int rc;
- rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
+ if (llcwrk->link->state == SMC_LNK_INACTIVE)
+ goto out;
+ rc = smc_llc_add_pending_send(llcwrk->link, &wr_buf, &pend);
if (rc)
- return rc;
- memcpy(wr_buf, llcbuf, llclen);
- /* send llc message */
- rc = smc_wr_tx_send(link, pend);
- return rc;
+ goto out;
+ memcpy(wr_buf, &llcwrk->llcbuf, llcwrk->llclen);
+ smc_wr_tx_send(llcwrk->link, pend);
+out:
+ kfree(llcwrk);
+}
+
+/* copy llcbuf and schedule an llc send on link */
+static int smc_llc_send_message(struct smc_link *link, void *llcbuf, int llclen)
+{
+ struct smc_llc_send_work *wrk = kmalloc(sizeof(*wrk), GFP_ATOMIC);
+
+ if (!wrk)
+ return -ENOMEM;
+ INIT_WORK(&wrk->work, smc_llc_send_message_work);
+ wrk->link = link;
+ wrk->llclen = llclen;
+ memcpy(&wrk->llcbuf, llcbuf, llclen);
+ queue_work(link->llc_wq, &wrk->work);
+ return 0;
}
/********************************* receive ***********************************/
@@ -381,17 +424,18 @@ static void smc_llc_rx_add_link(struct smc_link *link,
}
if (lgr->role == SMC_SERV) {
- smc_llc_send_add_link(link,
+ smc_llc_prep_add_link(llc, link,
link->smcibdev->mac[link->ibport - 1],
&link->smcibdev->gid[link->ibport - 1],
SMC_LLC_REQ);
} else {
- smc_llc_send_add_link(link,
+ smc_llc_prep_add_link(llc, link,
link->smcibdev->mac[link->ibport - 1],
&link->smcibdev->gid[link->ibport - 1],
SMC_LLC_RESP);
}
+ smc_llc_send_message(link, llc, sizeof(*llc));
}
}
@@ -407,9 +451,11 @@ static void smc_llc_rx_delete_link(struct smc_link *link,
} else {
if (lgr->role == SMC_SERV) {
smc_lgr_forget(lgr);
- smc_llc_send_delete_link(link, SMC_LLC_REQ);
+ smc_llc_prep_delete_link(llc, link, SMC_LLC_REQ);
+ smc_llc_send_message(link, llc, sizeof(*llc));
} else {
- smc_llc_send_delete_link(link, SMC_LLC_RESP);
+ smc_llc_prep_delete_link(llc, link, SMC_LLC_RESP);
+ smc_llc_send_message(link, llc, sizeof(*llc));
smc_lgr_terminate(lgr);
}
}
@@ -559,11 +605,19 @@ static void smc_llc_testlink_work(struct work_struct *work)
}
next_interval = link->llc_testlink_time;
out:
- schedule_delayed_work(&link->llc_testlink_wrk, next_interval);
+ queue_delayed_work(link->llc_wq, &link->llc_testlink_wrk,
+ next_interval);
}
-void smc_llc_link_init(struct smc_link *link)
+int smc_llc_link_init(struct smc_link *link)
{
+ struct smc_link_group *lgr = container_of(link, struct smc_link_group,
+ lnk[SMC_SINGLE_LINK]);
+ link->llc_wq = alloc_ordered_workqueue("llc_wq-%x:%x)", WQ_MEM_RECLAIM,
+ *((u32 *)lgr->id),
+ link->link_id);
+ if (!link->llc_wq)
+ return -ENOMEM;
init_completion(&link->llc_confirm);
init_completion(&link->llc_confirm_resp);
init_completion(&link->llc_add);
@@ -571,6 +625,7 @@ void smc_llc_link_init(struct smc_link *link)
init_completion(&link->llc_confirm_rkey);
init_completion(&link->llc_testlink_resp);
INIT_DELAYED_WORK(&link->llc_testlink_wrk, smc_llc_testlink_work);
+ return 0;
}
void smc_llc_link_active(struct smc_link *link, int testlink_time)
@@ -578,8 +633,8 @@ void smc_llc_link_active(struct smc_link *link, int testlink_time)
link->state = SMC_LNK_ACTIVE;
if (testlink_time) {
link->llc_testlink_time = testlink_time * HZ;
- schedule_delayed_work(&link->llc_testlink_wrk,
- link->llc_testlink_time);
+ queue_delayed_work(link->llc_wq, &link->llc_testlink_wrk,
+ link->llc_testlink_time);
}
}
@@ -591,9 +646,10 @@ void smc_llc_link_inactive(struct smc_link *link)
}
/* called in worker context */
-void smc_llc_link_flush(struct smc_link *link)
+void smc_llc_link_clear(struct smc_link *link)
{
- cancel_delayed_work_sync(&link->llc_testlink_wrk);
+ flush_workqueue(link->llc_wq);
+ destroy_workqueue(link->llc_wq);
}
/* register a new rtoken at the remote peer */
diff --git a/net/smc/smc_llc.h b/net/smc/smc_llc.h
index d7f52a60c8d6..65c8645e96a1 100644
--- a/net/smc/smc_llc.h
+++ b/net/smc/smc_llc.h
@@ -42,10 +42,10 @@ int smc_llc_send_add_link(struct smc_link *link, u8 mac[], union ib_gid *gid,
enum smc_llc_reqresp reqresp);
int smc_llc_send_delete_link(struct smc_link *link,
enum smc_llc_reqresp reqresp);
-void smc_llc_link_init(struct smc_link *link);
+int smc_llc_link_init(struct smc_link *link);
void smc_llc_link_active(struct smc_link *link, int testlink_time);
void smc_llc_link_inactive(struct smc_link *link);
-void smc_llc_link_flush(struct smc_link *link);
+void smc_llc_link_clear(struct smc_link *link);
int smc_llc_do_confirm_rkey(struct smc_link *link,
struct smc_buf_desc *rmb_desc);
int smc_llc_init(void) __init;
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 05/10] net/smc: move link llc initialization to llc layer
From: Ursula Braun @ 2018-05-15 15:04 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Move the llc layer specific initialization and cleanup out of smc_core.c
into smc_llc.c (smc_llc_link_init and smc_llc_link_clear). Move all
initialization of a link into the new init function.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/smc_core.c | 6 +-----
net/smc/smc_llc.c | 11 ++++++++++-
net/smc/smc_llc.h | 1 +
3 files changed, 12 insertions(+), 6 deletions(-)
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index 29ae077b0ec9..ba7dae819ac8 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -194,6 +194,7 @@ static int smc_lgr_create(struct smc_sock *smc,
smc_ib_setup_per_ibdev(smcibdev);
get_random_bytes(rndvec, sizeof(rndvec));
lnk->psn_initial = rndvec[0] + (rndvec[1] << 8) + (rndvec[2] << 16);
+ smc_llc_link_init(lnk);
rc = smc_wr_alloc_link_mem(lnk);
if (rc)
goto free_lgr;
@@ -206,11 +207,6 @@ static int smc_lgr_create(struct smc_sock *smc,
rc = smc_wr_create_link(lnk);
if (rc)
goto destroy_qp;
- init_completion(&lnk->llc_confirm);
- init_completion(&lnk->llc_confirm_resp);
- init_completion(&lnk->llc_add);
- init_completion(&lnk->llc_add_resp);
- init_completion(&lnk->llc_confirm_rkey);
smc->conn.lgr = lgr;
rwlock_init(&lgr->conns_lock);
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 229e967c7940..5dbeced6cea5 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -562,10 +562,19 @@ static void smc_llc_testlink_work(struct work_struct *work)
schedule_delayed_work(&link->llc_testlink_wrk, next_interval);
}
-void smc_llc_link_active(struct smc_link *link, int testlink_time)
+void smc_llc_link_init(struct smc_link *link)
{
+ init_completion(&link->llc_confirm);
+ init_completion(&link->llc_confirm_resp);
+ init_completion(&link->llc_add);
+ init_completion(&link->llc_add_resp);
+ init_completion(&link->llc_confirm_rkey);
init_completion(&link->llc_testlink_resp);
INIT_DELAYED_WORK(&link->llc_testlink_wrk, smc_llc_testlink_work);
+}
+
+void smc_llc_link_active(struct smc_link *link, int testlink_time)
+{
link->state = SMC_LNK_ACTIVE;
if (testlink_time) {
link->llc_testlink_time = testlink_time * HZ;
diff --git a/net/smc/smc_llc.h b/net/smc/smc_llc.h
index fb137f24aa6b..d7f52a60c8d6 100644
--- a/net/smc/smc_llc.h
+++ b/net/smc/smc_llc.h
@@ -42,6 +42,7 @@ int smc_llc_send_add_link(struct smc_link *link, u8 mac[], union ib_gid *gid,
enum smc_llc_reqresp reqresp);
int smc_llc_send_delete_link(struct smc_link *link,
enum smc_llc_reqresp reqresp);
+void smc_llc_link_init(struct smc_link *link);
void smc_llc_link_active(struct smc_link *link, int testlink_time);
void smc_llc_link_inactive(struct smc_link *link);
void smc_llc_link_flush(struct smc_link *link);
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 08/10] net/smc: set link inactive before calling smc_lgr_free()
From: Ursula Braun @ 2018-05-15 15:05 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Before smc_lgr_free() is called the link must be set inactive by calling
smc_llc_link_inactive().
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/af_smc.c | 1 +
net/smc/smc_core.c | 5 ++++-
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index ecf9ba68501b..d15762b057c0 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -1644,6 +1644,7 @@ static void __exit smc_exit(void)
spin_unlock_bh(&smc_lgr_list.lock);
list_for_each_entry_safe(lgr, lg, &lgr_freeing_list, list) {
list_del_init(&lgr->list);
+ smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
cancel_delayed_work_sync(&lgr->free_work);
smc_lgr_free(lgr); /* free link group */
}
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index 4523bbad8a15..a48ee00b65ff 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -148,8 +148,11 @@ static void smc_lgr_free_work(struct work_struct *work)
list_del_init(&lgr->list); /* remove from smc_lgr_list */
free:
spin_unlock_bh(&smc_lgr_list.lock);
- if (!delayed_work_pending(&lgr->free_work))
+ if (!delayed_work_pending(&lgr->free_work)) {
+ if (lgr->lnk[SMC_SINGLE_LINK].state != SMC_LNK_INACTIVE)
+ smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
smc_lgr_free(lgr);
+ }
}
/* create a new SMC link group */
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 04/10] net/smc: simplify test_link function usage
From: Ursula Braun @ 2018-05-15 15:04 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Make smc_llc_send_test_link() static and remove it from the header file.
And to send a test_link response set the response flag and send the
message back as-is, without using smc_llc_send_test_link(). And because
smc_llc_send_test_link() must no longer send responses, remove the
response flag handling from the function.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/smc_llc.c | 12 +++++-------
net/smc/smc_llc.h | 2 --
2 files changed, 5 insertions(+), 9 deletions(-)
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 29f59b47b559..229e967c7940 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -296,9 +296,8 @@ int smc_llc_send_delete_link(struct smc_link *link,
return rc;
}
-/* send LLC test link request or response */
-int smc_llc_send_test_link(struct smc_link *link, u8 user_data[16],
- enum smc_llc_reqresp reqresp)
+/* send LLC test link request */
+static int smc_llc_send_test_link(struct smc_link *link, u8 user_data[16])
{
struct smc_llc_msg_test_link *testllc;
struct smc_wr_tx_pend_priv *pend;
@@ -312,8 +311,6 @@ int smc_llc_send_test_link(struct smc_link *link, u8 user_data[16],
memset(testllc, 0, sizeof(*testllc));
testllc->hd.common.type = SMC_LLC_TEST_LINK;
testllc->hd.length = sizeof(struct smc_llc_msg_test_link);
- if (reqresp == SMC_LLC_RESP)
- testllc->hd.flags |= SMC_LLC_FLAG_RESP;
memcpy(testllc->user_data, user_data, sizeof(testllc->user_data));
/* send llc message */
rc = smc_wr_tx_send(link, pend);
@@ -425,7 +422,8 @@ static void smc_llc_rx_test_link(struct smc_link *link,
if (link->state == SMC_LNK_ACTIVE)
complete(&link->llc_testlink_resp);
} else {
- smc_llc_send_test_link(link, llc->user_data, SMC_LLC_RESP);
+ llc->hd.flags |= SMC_LLC_FLAG_RESP;
+ smc_llc_send_message(link, llc, sizeof(*llc));
}
}
@@ -551,7 +549,7 @@ static void smc_llc_testlink_work(struct work_struct *work)
goto out;
}
reinit_completion(&link->llc_testlink_resp);
- smc_llc_send_test_link(link, user_data, SMC_LLC_REQ);
+ smc_llc_send_test_link(link, user_data);
/* receive TEST LINK response over RoCE fabric */
rc = wait_for_completion_interruptible_timeout(&link->llc_testlink_resp,
SMC_LLC_WAIT_TIME);
diff --git a/net/smc/smc_llc.h b/net/smc/smc_llc.h
index 728823dd8ae5..fb137f24aa6b 100644
--- a/net/smc/smc_llc.h
+++ b/net/smc/smc_llc.h
@@ -42,8 +42,6 @@ int smc_llc_send_add_link(struct smc_link *link, u8 mac[], union ib_gid *gid,
enum smc_llc_reqresp reqresp);
int smc_llc_send_delete_link(struct smc_link *link,
enum smc_llc_reqresp reqresp);
-int smc_llc_send_test_link(struct smc_link *lnk, u8 user_data[16],
- enum smc_llc_reqresp reqresp);
void smc_llc_link_active(struct smc_link *link, int testlink_time);
void smc_llc_link_inactive(struct smc_link *link);
void smc_llc_link_flush(struct smc_link *link);
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 09/10] net/smc: drop messages when link state is inactive
From: Ursula Braun @ 2018-05-15 15:05 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Drop incoming messages when the link is flagged as inactive.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/smc_llc.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 5d7926cc472c..5800a6b43d83 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -548,6 +548,8 @@ static void smc_llc_rx_handler(struct ib_wc *wc, void *buf)
return; /* short message */
if (llc->raw.hdr.length != sizeof(*llc))
return; /* invalid message */
+ if (link->state == SMC_LNK_INACTIVE)
+ return; /* link not active, drop msg */
switch (llc->raw.hdr.common.type) {
case SMC_LLC_TEST_LINK:
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 07/10] net/smc: handle all error codes from smc_conn_create()
From: Ursula Braun @ 2018-05-15 15:05 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Always set a reason_code when smc_conn_create() returns an error code.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/af_smc.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 397ba2182453..ecf9ba68501b 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -463,6 +463,8 @@ static int smc_connect_rdma(struct smc_sock *smc)
reason_code = SMC_CLC_DECL_MEM;/* insufficient memory*/
else if (rc == -ENOLINK)
reason_code = SMC_CLC_DECL_SYNCERR; /* synchr. error */
+ else
+ reason_code = SMC_CLC_DECL_INTERR; /* other error */
goto decline_rdma_unlock;
}
link = &smc->conn.lgr->lnk[SMC_SINGLE_LINK];
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 10/10] net/smc: check for pending termination
From: Ursula Braun @ 2018-05-15 15:05 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Avoid to run the processing in smc_lgr_terminate() more than once,
remember when the link group termination is triggered.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/smc_clc.c | 2 +-
net/smc/smc_core.c | 5 ++++-
net/smc/smc_core.h | 3 ++-
3 files changed, 7 insertions(+), 3 deletions(-)
diff --git a/net/smc/smc_clc.c b/net/smc/smc_clc.c
index 3a988c22f627..236cb3f12c71 100644
--- a/net/smc/smc_clc.c
+++ b/net/smc/smc_clc.c
@@ -316,7 +316,7 @@ int smc_clc_wait_msg(struct smc_sock *smc, void *buf, int buflen,
if (clcm->type == SMC_CLC_DECLINE) {
reason_code = SMC_CLC_DECL_REPLY;
if (((struct smc_clc_msg_decline *)buf)->hdr.flag) {
- smc->conn.lgr->sync_err = true;
+ smc->conn.lgr->sync_err = 1;
smc_lgr_terminate(smc->conn.lgr);
}
}
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index a48ee00b65ff..08c05cd0bbae 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -172,7 +172,7 @@ static int smc_lgr_create(struct smc_sock *smc,
goto out;
}
lgr->role = smc->listen_smc ? SMC_SERV : SMC_CLNT;
- lgr->sync_err = false;
+ lgr->sync_err = 0;
memcpy(lgr->peer_systemid, peer_systemid, SMC_SYSTEMID_LEN);
lgr->vlan_id = vlan_id;
rwlock_init(&lgr->sndbufs_lock);
@@ -352,6 +352,9 @@ void smc_lgr_terminate(struct smc_link_group *lgr)
struct smc_sock *smc;
struct rb_node *node;
+ if (lgr->terminating)
+ return; /* lgr already terminating */
+ lgr->terminating = 1;
smc_lgr_forget(lgr);
smc_llc_link_inactive(&lgr->lnk[SMC_SINGLE_LINK]);
diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h
index d1753850684b..845dc073de13 100644
--- a/net/smc/smc_core.h
+++ b/net/smc/smc_core.h
@@ -166,7 +166,8 @@ struct smc_link_group {
u8 id[SMC_LGR_ID_SIZE]; /* unique lgr id */
struct delayed_work free_work; /* delayed freeing of an lgr */
- bool sync_err; /* lgr no longer fits to peer */
+ u8 sync_err : 1; /* lgr no longer fits to peer */
+ u8 terminating : 1;/* lgr is terminating */
};
/* Find the connection associated with the given alert token in the link group.
--
2.16.3
^ permalink raw reply related
* [PATCH net-next 03/10] net/smc: remove unnecessary cast
From: Ursula Braun @ 2018-05-15 15:04 UTC (permalink / raw)
To: davem; +Cc: netdev, linux-s390, schwidefsky, heiko.carstens, raspl, ubraun
In-Reply-To: <20180515150503.47265-1-ubraun@linux.ibm.com>
From: Karsten Graul <kgraul@linux.ibm.com>
Remove an unneeded (void *) cast from the calls to
smc_llc_send_message(). No functional changes.
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Ursula Braun <ubraun@linux.ibm.com>
---
net/smc/smc_llc.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index b118f80b9c41..29f59b47b559 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -451,7 +451,7 @@ static void smc_llc_rx_confirm_rkey(struct smc_link *link,
llc->hd.flags |= SMC_LLC_FLAG_RESP;
if (rc < 0)
llc->hd.flags |= SMC_LLC_FLAG_RKEY_NEG;
- smc_llc_send_message(link, (void *)llc, sizeof(*llc));
+ smc_llc_send_message(link, llc, sizeof(*llc));
}
}
@@ -463,7 +463,7 @@ static void smc_llc_rx_confirm_rkey_cont(struct smc_link *link,
} else {
/* ignore rtokens for other links, we have only one link */
llc->hd.flags |= SMC_LLC_FLAG_RESP;
- smc_llc_send_message(link, (void *)llc, sizeof(*llc));
+ smc_llc_send_message(link, llc, sizeof(*llc));
}
}
@@ -491,7 +491,7 @@ static void smc_llc_rx_delete_rkey(struct smc_link *link,
}
llc->hd.flags |= SMC_LLC_FLAG_RESP;
- smc_llc_send_message(link, (void *)llc, sizeof(*llc));
+ smc_llc_send_message(link, llc, sizeof(*llc));
}
}
--
2.16.3
^ permalink raw reply related
* [PATCH net-next v11 0/7] sched: Add Common Applications Kept Enhanced (cake) qdisc
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
This patch series adds the CAKE qdisc, and has been split up to ease
review.
I have attempted to split out each configurable feature into its own patch.
The first commit adds the base shaper and packet scheduler, while
subsequent commits add the optional features. The full userspace API and
most data structures are included in this commit, but options not
understood in the base version will be ignored.
The result of applying the entire series is identical to the out of tree
version that have seen extensive testing in previous deployments, most
notably as an out of tree patch to OpenWrt. However, note that I have only
compile tested the individual patches; so the whole series should be
considered as a unit.
---
Changelog
v11:
- Fix overhead compensation calculation for GSO packets
- Change configured rate to be u64 (I ran out of bits before I ran out
of CPU when testing the effects of the above)
v10:
- Christmas tree gardening (fix variable declarations to be in reverse
line length order)
v9:
- Remove duplicated checks around kvfree() and just call it
unconditionally.
- Don't pass __GFP_NOWARN when allocating memory
- Move options in cake_dump() that are related to optional features to
later patches implementing the features.
- Support attaching filters to the qdisc and use the classification
result to select flow queue.
- Support overriding diffserv priority tin from skb->priority
v8:
- Remove inline keyword from function definitions
- Simplify ACK filter; remove the complex state handling to make the
logic easier to follow. This will potentially be a bit less efficient,
but I have not been able to measure a difference.
v7:
- Split up patch into a series to ease review.
- Constify the ACK filter.
v6:
- Fix 6in4 encapsulation checks in ACK filter code
- Checkpatch fixes
v5:
- Refactor ACK filter code and hopefully fix the safety issues
properly this time.
v4:
- Only split GSO packets if shaping at speeds <= 1Gbps
- Fix overhead calculation code to also work for GSO packets
- Don't re-implement kvzalloc()
- Remove local header include from out-of-tree build (fixes kbuild-bot
complaint).
- Several fixes to the ACK filter:
- Check pskb_may_pull() before deref of transport headers.
- Don't run ACK filter logic on split GSO packets
- Fix TCP sequence number compare to deal with wraparounds
v3:
- Use IS_REACHABLE() macro to fix compilation when sch_cake is
built-in and conntrack is a module.
- Switch the stats output to use nested netlink attributes instead
of a versioned struct.
- Remove GPL boilerplate.
- Fix array initialisation style.
v2:
- Fix kbuild test bot complaint
- Clean up the netlink ABI
- Fix checkpatch complaints
- A few tweaks to the behaviour of cake based on testing carried out
while writing the paper.
---
Toke Høiland-Jørgensen (7):
sched: Add Common Applications Kept Enhanced (cake) qdisc
sch_cake: Add ingress mode
sch_cake: Add optional ACK filter
sch_cake: Add NAT awareness to packet classifier
sch_cake: Add DiffServ handling
sch_cake: Add overhead compensation support to the rate shaper
sch_cake: Conditionally split GSO segments
include/uapi/linux/pkt_sched.h | 105 ++
net/sched/Kconfig | 11
net/sched/Makefile | 1
net/sched/sch_cake.c | 2692 ++++++++++++++++++++++++++++++++++++++++
4 files changed, 2809 insertions(+)
create mode 100644 net/sched/sch_cake.c
^ permalink raw reply
* [PATCH net-next v11 3/7] sch_cake: Add optional ACK filter
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
In-Reply-To: <152639705919.2525.3718182212727523960.stgit@alrua-kau>
The ACK filter is an optional feature of CAKE which is designed to improve
performance on links with very asymmetrical rate limits. On such links
(which are unfortunately quite prevalent, especially for DSL and cable
subscribers), the downstream throughput can be limited by the number of
ACKs capable of being transmitted in the *upstream* direction.
Filtering ACKs can, in general, have adverse effects on TCP performance
because it interferes with ACK clocking (especially in slow start), and it
reduces the flow's resiliency to ACKs being dropped further along the path.
To alleviate these drawbacks, the ACK filter in CAKE tries its best to
always keep enough ACKs queued to ensure forward progress in the TCP flow
being filtered. It does this by only filtering redundant ACKs. In its
default 'conservative' mode, the filter will always keep at least two
redundant ACKs in the queue, while in 'aggressive' mode, it will filter
down to a single ACK.
The ACK filter works by inspecting the per-flow queue on every packet
enqueue. Starting at the head of the queue, the filter looks for another
eligible packet to drop (so the ACK being dropped is always closer to the
head of the queue than the packet being enqueued). An ACK is eligible only
if it ACKs *fewer* cumulative bytes than the new packet being enqueued.
This prevents duplicate ACKs from being filtered (unless there is also SACK
options present), to avoid interfering with retransmission logic. In
aggressive mode, an eligible packet is always dropped, while in
conservative mode, at least two ACKs are kept in the queue. Only pure ACKs
(with no data segments) are considered eligible for dropping, but when an
ACK with data segments is enqueued, this can cause another pure ACK to
become eligible for dropping.
The approach described above ensures that this ACK filter avoids most of
the drawbacks of a naive filtering mechanism that only keeps flow state but
does not inspect the queue. This is the rationale for including the ACK
filter in CAKE itself rather than as separate module (as the TC filter, for
instance).
Our performance evaluation has shown that on a 30/1 Mbps link with a
bidirectional traffic test (RRUL), turning on the ACK filter on the
upstream link improves downstream throughput by ~20% (both modes) and
upstream throughput by ~12% in conservative mode and ~40% in aggressive
mode, at the cost of ~5ms of inter-flow latency due to the increased
congestion.
In *really* pathological cases, the effect can be a lot more; for instance,
the ACK filter increases the achievable downstream throughput on a link
with 100 Kbps in the upstream direction by an order of magnitude (from ~2.5
Mbps to ~25 Mbps).
Finally, even though we consider the ACK filter to be safer than most, we
do not recommend turning it on everywhere: on more symmetrical link
bandwidths the effect is negligible at best.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
net/sched/sch_cake.c | 260 ++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 258 insertions(+), 2 deletions(-)
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 8d0823d6d8dd..371c888cb982 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -763,6 +763,239 @@ static void flow_queue_add(struct cake_flow *flow, struct sk_buff *skb)
skb->next = NULL;
}
+static struct iphdr *cake_get_iphdr(const struct sk_buff *skb,
+ struct ipv6hdr *buf)
+{
+ unsigned int offset = skb_network_offset(skb);
+ struct iphdr *iph;
+
+ iph = skb_header_pointer(skb, offset, sizeof(struct iphdr), buf);
+
+ if (!iph)
+ return NULL;
+
+ if (iph->version == 4 && iph->protocol == IPPROTO_IPV6)
+ return skb_header_pointer(skb, offset + iph->ihl * 4,
+ sizeof(struct ipv6hdr), buf);
+
+ else if (iph->version == 4)
+ return iph;
+
+ else if (iph->version == 6)
+ return skb_header_pointer(skb, offset, sizeof(struct ipv6hdr),
+ buf);
+
+ return NULL;
+}
+
+static struct tcphdr *cake_get_tcphdr(const struct sk_buff *skb,
+ void *buf, unsigned int bufsize)
+{
+ unsigned int offset = skb_network_offset(skb);
+ const struct ipv6hdr *ipv6h;
+ const struct tcphdr *tcph;
+ const struct iphdr *iph;
+ struct ipv6hdr _ipv6h;
+ struct tcphdr _tcph;
+
+ ipv6h = skb_header_pointer(skb, offset, sizeof(_ipv6h), &_ipv6h);
+
+ if (!ipv6h)
+ return NULL;
+
+ if (ipv6h->version == 4) {
+ iph = (struct iphdr *)ipv6h;
+ offset += iph->ihl * 4;
+
+ /* special-case 6in4 tunnelling, as that is a common way to get
+ * v6 connectivity in the home
+ */
+ if (iph->protocol == IPPROTO_IPV6) {
+ ipv6h = skb_header_pointer(skb, offset,
+ sizeof(_ipv6h), &_ipv6h);
+
+ if (!ipv6h || ipv6h->nexthdr != IPPROTO_TCP)
+ return NULL;
+
+ offset += sizeof(struct ipv6hdr);
+
+ } else if (iph->protocol != IPPROTO_TCP) {
+ return NULL;
+ }
+
+ } else if (ipv6h->version == 6) {
+ if (ipv6h->nexthdr != IPPROTO_TCP)
+ return NULL;
+
+ offset += sizeof(struct ipv6hdr);
+ } else {
+ return NULL;
+ }
+
+ tcph = skb_header_pointer(skb, offset, sizeof(_tcph), &_tcph);
+ if (!tcph)
+ return NULL;
+
+ return skb_header_pointer(skb, offset,
+ min(__tcp_hdrlen(tcph), bufsize), buf);
+}
+
+static bool cake_tcph_is_sack(const struct tcphdr *tcph)
+{
+ /* inspired by tcp_parse_options in tcp_input.c */
+ int length = __tcp_hdrlen(tcph) - sizeof(struct tcphdr);
+ const u8 *ptr = (const u8 *)(tcph + 1);
+
+ while (length > 0) {
+ int opcode = *ptr++;
+ int opsize;
+
+ if (opcode == TCPOPT_EOL)
+ break;
+ if (opcode == TCPOPT_NOP) {
+ length--;
+ continue;
+ }
+ opsize = *ptr++;
+ if (opsize < 2 || opsize > length)
+ break;
+ if (opcode == TCPOPT_SACK)
+ return true;
+ ptr += opsize - 2;
+ length -= opsize;
+ }
+
+ return false;
+}
+
+static struct sk_buff *cake_ack_filter(struct cake_sched_data *q,
+ struct cake_flow *flow)
+{
+ bool aggressive = q->ack_filter == CAKE_ACK_AGGRESSIVE;
+ struct sk_buff *elig_ack = NULL, *elig_ack_prev = NULL;
+ struct sk_buff *skb_check, *skb_prev = NULL;
+ const struct ipv6hdr *ipv6h, *ipv6h_check;
+ const struct tcphdr *tcph, *tcph_check;
+ const struct iphdr *iph, *iph_check;
+ struct ipv6hdr _iph, _iph_check;
+ const struct sk_buff *skb;
+ struct tcphdr _tcph_check;
+ int seglen, num_found = 0;
+ unsigned char _tcph[64]; /* need to hold maximum hdr size */
+
+ /* no other possible ACKs to filter */
+ if (flow->head == flow->tail)
+ return NULL;
+
+ skb = flow->tail;
+ tcph = cake_get_tcphdr(skb, _tcph, sizeof(_tcph));
+ iph = cake_get_iphdr(skb, &_iph);
+ if (!tcph)
+ return NULL;
+
+ /* the 'triggering' packet need only have the ACK flag set.
+ * also check that SYN is not set, as there won't be any previous ACKs.
+ */
+ if ((tcp_flag_word(tcph) &
+ (TCP_FLAG_ACK | TCP_FLAG_SYN)) != TCP_FLAG_ACK)
+ return NULL;
+
+ /* the 'triggering' ACK is at the tail of the queue, we have already
+ * returned if it is the only packet in the flow. loop through the rest
+ * of the queue looking for pure ACKs with the same 5-tuple as the
+ * triggering one.
+ */
+ for (skb_check = flow->head;
+ skb_check && skb_check != skb;
+ skb_prev = skb_check, skb_check = skb_check->next) {
+ iph_check = cake_get_iphdr(skb_check, &_iph_check);
+ tcph_check = cake_get_tcphdr(skb_check, &_tcph_check,
+ sizeof(_tcph_check));
+
+ /* only TCP packets with matching 5-tuple are eligible */
+ if (!tcph_check || iph->version != iph_check->version ||
+ tcph_check->source != tcph->source ||
+ tcph_check->dest != tcph->dest)
+ continue;
+
+ if (iph_check->version == 4) {
+ if (iph_check->saddr != iph->saddr ||
+ iph_check->daddr != iph->daddr)
+ continue;
+
+ seglen = ntohs(iph_check->tot_len) -
+ (4 * iph_check->ihl);
+ } else if (iph_check->version == 6) {
+ ipv6h = (struct ipv6hdr *)iph;
+ ipv6h_check = (struct ipv6hdr *)iph_check;
+
+ if (ipv6_addr_cmp(&ipv6h_check->saddr, &ipv6h->saddr) ||
+ ipv6_addr_cmp(&ipv6h_check->daddr, &ipv6h->daddr))
+ continue;
+
+ seglen = ntohs(ipv6h_check->payload_len);
+ } else {
+ WARN_ON(1); /* shouldn't happen */
+ continue;
+ }
+
+ /* stricter criteria apply to ACKs that we may filter
+ * 3 reserved flags must be unset to avoid future breakage
+ * ECE/CWR/NS can be safely ignored
+ * ACK must be set
+ * All other flags URG/PSH/RST/SYN/FIN must be unset
+ * 0x0FFF0000 = all TCP flags (confirm ACK=1, others zero)
+ * 0x01C00000 = NS/CWR/ECE (safe to ignore)
+ * 0x0E3F0000 = 0x0FFF0000 & ~0x01C00000
+ * must be 'pure' ACK, contain zero bytes of segment data
+ * options are ignored
+ */
+ if (((tcp_flag_word(tcph_check) &
+ cpu_to_be32(0x0E3F0000)) != TCP_FLAG_ACK) ||
+ ((seglen - __tcp_hdrlen(tcph_check)) != 0))
+ continue;
+
+ /* The triggering packet must ACK more data than the ACK under
+ * consideration, either because is has a strictly higher ACK
+ * sequence number or because it is a SACK
+ */
+ if ((ntohl(tcph_check->ack_seq) == ntohl(tcph->ack_seq) &&
+ !cake_tcph_is_sack(tcph)) ||
+ (int32_t)(ntohl(tcph_check->ack_seq) -
+ ntohl(tcph->ack_seq)) > 0)
+ continue;
+
+ /* At this point we have found an eligible pure ACK to drop; if
+ * we are in aggressive mode, we are done. Otherwise, keep
+ * searching unless this is the second eligible ACK we
+ * found.
+ *
+ * Since we want to drop ACK closest to the head of the queue,
+ * save the first eligible ACK we find, even if we need to loop
+ * again.
+ */
+ if (!elig_ack) {
+ elig_ack = skb_check;
+ elig_ack_prev = skb_prev;
+ }
+
+ if (num_found++ > 0 || aggressive)
+ goto found;
+ }
+
+ return NULL;
+
+found:
+ if (elig_ack_prev)
+ elig_ack_prev->next = elig_ack->next;
+ else
+ flow->head = elig_ack->next;
+
+ elig_ack->next = NULL;
+
+ return elig_ack;
+}
+
static cobalt_time_t cake_ewma(cobalt_time_t avg, cobalt_time_t sample,
u32 shift)
{
@@ -936,6 +1169,7 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
int len = qdisc_pkt_len(skb);
u64 now = cobalt_get_time();
int uninitialized_var(ret);
+ struct sk_buff *ack = NULL;
struct cake_tin_data *b;
struct cake_flow *flow;
u32 idx, tin;
@@ -979,8 +1213,24 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
cobalt_set_enqueue_time(skb, now);
flow_queue_add(flow, skb);
- sch->q.qlen++;
- q->buffer_used += skb->truesize;
+ if (q->ack_filter)
+ ack = cake_ack_filter(q, flow);
+
+ if (ack) {
+ b->ack_drops++;
+ sch->qstats.drops++;
+ b->bytes += qdisc_pkt_len(ack);
+ len -= qdisc_pkt_len(ack);
+ q->buffer_used += skb->truesize - ack->truesize;
+ if (q->rate_flags & CAKE_FLAG_INGRESS)
+ cake_advance_shaper(q, b, ack, now, true);
+
+ qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(ack));
+ consume_skb(ack);
+ } else {
+ sch->q.qlen++;
+ q->buffer_used += skb->truesize;
+ }
/* stats */
b->packets++;
@@ -1493,6 +1743,9 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
q->rate_flags &= ~CAKE_FLAG_INGRESS;
}
+ if (tb[TCA_CAKE_ACK_FILTER])
+ q->ack_filter = nla_get_u32(tb[TCA_CAKE_ACK_FILTER]);
+
if (tb[TCA_CAKE_MEMORY])
q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
@@ -1624,6 +1877,9 @@ static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
!!(q->rate_flags & CAKE_FLAG_INGRESS)))
goto nla_put_failure;
+ if (nla_put_u32(skb, TCA_CAKE_ACK_FILTER, q->ack_filter))
+ goto nla_put_failure;
+
return nla_nest_end(skb, opts);
nla_put_failure:
^ permalink raw reply related
* [PATCH net-next v11 4/7] sch_cake: Add NAT awareness to packet classifier
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
In-Reply-To: <152639705919.2525.3718182212727523960.stgit@alrua-kau>
When CAKE is deployed on a gateway that also performs NAT (which is a
common deployment mode), the host fairness mechanism cannot distinguish
internal hosts from each other, and so fails to work correctly.
To fix this, we add an optional NAT awareness mode, which will query the
kernel conntrack mechanism to obtain the pre-NAT addresses for each packet
and use that in the flow and host hashing.
When the shaper is enabled and the host is already performing NAT, the cost
of this lookup is negligible. However, in unlimited mode with no NAT being
performed, there is a significant CPU cost at higher bandwidths. For this
reason, the feature is turned off by default.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
net/sched/sch_cake.c | 73 ++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 73 insertions(+)
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 371c888cb982..14622c1d970e 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -71,6 +71,12 @@
#include <net/tcp.h>
#include <net/flow_dissector.h>
+#if IS_REACHABLE(CONFIG_NF_CONNTRACK)
+#include <net/netfilter/nf_conntrack_core.h>
+#include <net/netfilter/nf_conntrack_zones.h>
+#include <net/netfilter/nf_conntrack.h>
+#endif
+
#define CAKE_SET_WAYS (8)
#define CAKE_MAX_TINS (8)
#define CAKE_QUEUES (1024)
@@ -522,6 +528,60 @@ static bool cobalt_should_drop(struct cobalt_vars *vars,
return drop;
}
+#if IS_REACHABLE(CONFIG_NF_CONNTRACK)
+
+static void cake_update_flowkeys(struct flow_keys *keys,
+ const struct sk_buff *skb)
+{
+ const struct nf_conntrack_tuple *tuple;
+ enum ip_conntrack_info ctinfo;
+ struct nf_conn *ct;
+ bool rev = false;
+
+ if (tc_skb_protocol(skb) != htons(ETH_P_IP))
+ return;
+
+ ct = nf_ct_get(skb, &ctinfo);
+ if (ct) {
+ tuple = nf_ct_tuple(ct, CTINFO2DIR(ctinfo));
+ } else {
+ const struct nf_conntrack_tuple_hash *hash;
+ struct nf_conntrack_tuple srctuple;
+
+ if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb),
+ NFPROTO_IPV4, dev_net(skb->dev),
+ &srctuple))
+ return;
+
+ hash = nf_conntrack_find_get(dev_net(skb->dev),
+ &nf_ct_zone_dflt,
+ &srctuple);
+ if (!hash)
+ return;
+
+ rev = true;
+ ct = nf_ct_tuplehash_to_ctrack(hash);
+ tuple = nf_ct_tuple(ct, !hash->tuple.dst.dir);
+ }
+
+ keys->addrs.v4addrs.src = rev ? tuple->dst.u3.ip : tuple->src.u3.ip;
+ keys->addrs.v4addrs.dst = rev ? tuple->src.u3.ip : tuple->dst.u3.ip;
+
+ if (keys->ports.ports) {
+ keys->ports.src = rev ? tuple->dst.u.all : tuple->src.u.all;
+ keys->ports.dst = rev ? tuple->src.u.all : tuple->dst.u.all;
+ }
+ if (rev)
+ nf_ct_put(ct);
+}
+#else
+static void cake_update_flowkeys(struct flow_keys *keys,
+ const struct sk_buff *skb)
+{
+ /* There is nothing we can do here without CONNTRACK */
+}
+#endif
+
/* Cake has several subtle multiple bit settings. In these cases you
* would be matching triple isolate mode as well.
*/
@@ -549,6 +609,9 @@ static u32 cake_hash(struct cake_tin_data *q, const struct sk_buff *skb,
skb_flow_dissect_flow_keys(skb, &keys,
FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
+ if (flow_mode & CAKE_FLOW_NAT_FLAG)
+ cake_update_flowkeys(&keys, skb);
+
/* flow_hash_from_keys() sorts the addresses by value, so we have
* to preserve their order in a separate data structure to treat
* src and dst host addresses as independently selectable.
@@ -1715,6 +1778,12 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
q->flow_mode = (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) &
CAKE_FLOW_MASK);
+ if (tb[TCA_CAKE_NAT]) {
+ q->flow_mode &= ~CAKE_FLOW_NAT_FLAG;
+ q->flow_mode |= CAKE_FLOW_NAT_FLAG *
+ !!nla_get_u32(tb[TCA_CAKE_NAT]);
+ }
+
if (tb[TCA_CAKE_RTT]) {
q->interval = nla_get_u32(tb[TCA_CAKE_RTT]);
@@ -1880,6 +1949,10 @@ static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
if (nla_put_u32(skb, TCA_CAKE_ACK_FILTER, q->ack_filter))
goto nla_put_failure;
+ if (nla_put_u32(skb, TCA_CAKE_NAT,
+ !!(q->flow_mode & CAKE_FLOW_NAT_FLAG)))
+ goto nla_put_failure;
+
return nla_nest_end(skb, opts);
nla_put_failure:
^ permalink raw reply related
* [PATCH net-next v11 1/7] sched: Add Common Applications Kept Enhanced (cake) qdisc
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
In-Reply-To: <152639705919.2525.3718182212727523960.stgit@alrua-kau>
sch_cake targets the home router use case and is intended to squeeze the
most bandwidth and latency out of even the slowest ISP links and routers,
while presenting an API simple enough that even an ISP can configure it.
Example of use on a cable ISP uplink:
tc qdisc add dev eth0 cake bandwidth 20Mbit nat docsis ack-filter
To shape a cable download link (ifb and tc-mirred setup elided)
tc qdisc add dev ifb0 cake bandwidth 200mbit nat docsis ingress wash
CAKE is filled with:
* A hybrid Codel/Blue AQM algorithm, "Cobalt", tied to an FQ_Codel
derived Flow Queuing system, which autoconfigures based on the bandwidth.
* A novel "triple-isolate" mode (the default) which balances per-host
and per-flow FQ even through NAT.
* An deficit based shaper, that can also be used in an unlimited mode.
* 8 way set associative hashing to reduce flow collisions to a minimum.
* A reasonable interpretation of various diffserv latency/loss tradeoffs.
* Support for zeroing diffserv markings for entering and exiting traffic.
* Support for interacting well with Docsis 3.0 shaper framing.
* Extensive support for DSL framing types.
* Support for ack filtering.
* Extensive statistics for measuring, loss, ecn markings, latency
variation.
A paper describing the design of CAKE is available at
https://arxiv.org/abs/1804.07617, and will be published at the 2018 IEEE
International Symposium on Local and Metropolitan Area Networks (LANMAN).
This patch adds the base shaper and packet scheduler, while subsequent
commits add the optional (configurable) features. The full userspace API
and most data structures are included in this commit, but options not
understood in the base version will be ignored.
Various versions baking have been available as an out of tree build for
kernel versions going back to 3.10, as the embedded router world has been
running a few years behind mainline Linux. A stable version has been
generally available on lede-17.01 and later.
sch_cake replaces a combination of iptables, tc filter, htb and fq_codel
in the sqm-scripts, with sane defaults and vastly simpler configuration.
CAKE's principal author is Jonathan Morton, with contributions from
Kevin Darbyshire-Bryant, Toke Høiland-Jørgensen, Sebastian Moeller,
Ryan Mounce, Guido Sarducci, Dean Scarff, Nils Andreas Svee, Dave Täht,
and Loganaden Velvindron.
Testing from Pete Heist, Georgios Amanakis, and the many other members of
the cake@lists.bufferbloat.net mailing list.
tc -s qdisc show dev eth2
qdisc cake 1: root refcnt 2 bandwidth 100Mbit diffserv3 triple-isolate rtt 100.0ms raw overhead 0
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0
memory used: 0b of 5000000b
capacity estimate: 100Mbit
min/max network layer size: 65535 / 0
min/max overhead-adjusted size: 65535 / 0
average network hdr offset: 0
Bulk Best Effort Voice
thresh 6250Kbit 100Mbit 25Mbit
target 5.0ms 5.0ms 5.0ms
interval 100.0ms 100.0ms 100.0ms
pk_delay 0us 0us 0us
av_delay 0us 0us 0us
sp_delay 0us 0us 0us
pkts 0 0 0
bytes 0 0 0
way_inds 0 0 0
way_miss 0 0 0
way_cols 0 0 0
drops 0 0 0
marks 0 0 0
ack_drop 0 0 0
sp_flows 0 0 0
bk_flows 0 0 0
un_flows 0 0 0
max_len 0 0 0
quantum 300 1514 762
Tested-by: Pete Heist <peteheist@gmail.com>
Tested-by: Georgios Amanakis <gamanakis@gmail.com>
Signed-off-by: Dave Taht <dave.taht@gmail.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
include/uapi/linux/pkt_sched.h | 105 ++
net/sched/Kconfig | 11
net/sched/Makefile | 1
net/sched/sch_cake.c | 1730 ++++++++++++++++++++++++++++++++++++++++
4 files changed, 1847 insertions(+)
create mode 100644 net/sched/sch_cake.c
diff --git a/include/uapi/linux/pkt_sched.h b/include/uapi/linux/pkt_sched.h
index 37b5096ae97b..883e84f008d7 100644
--- a/include/uapi/linux/pkt_sched.h
+++ b/include/uapi/linux/pkt_sched.h
@@ -934,4 +934,109 @@ enum {
#define TCA_CBS_MAX (__TCA_CBS_MAX - 1)
+/* CAKE */
+enum {
+ TCA_CAKE_UNSPEC,
+ TCA_CAKE_BASE_RATE64,
+ TCA_CAKE_DIFFSERV_MODE,
+ TCA_CAKE_ATM,
+ TCA_CAKE_FLOW_MODE,
+ TCA_CAKE_OVERHEAD,
+ TCA_CAKE_RTT,
+ TCA_CAKE_TARGET,
+ TCA_CAKE_AUTORATE,
+ TCA_CAKE_MEMORY,
+ TCA_CAKE_NAT,
+ TCA_CAKE_RAW,
+ TCA_CAKE_WASH,
+ TCA_CAKE_MPU,
+ TCA_CAKE_INGRESS,
+ TCA_CAKE_ACK_FILTER,
+ TCA_CAKE_SPLIT_GSO,
+ __TCA_CAKE_MAX
+};
+#define TCA_CAKE_MAX (__TCA_CAKE_MAX - 1)
+
+enum {
+ __TCA_CAKE_STATS_INVALID,
+ TCA_CAKE_STATS_CAPACITY_ESTIMATE64,
+ TCA_CAKE_STATS_MEMORY_LIMIT,
+ TCA_CAKE_STATS_MEMORY_USED,
+ TCA_CAKE_STATS_AVG_NETOFF,
+ TCA_CAKE_STATS_MIN_NETLEN,
+ TCA_CAKE_STATS_MAX_NETLEN,
+ TCA_CAKE_STATS_MIN_ADJLEN,
+ TCA_CAKE_STATS_MAX_ADJLEN,
+ TCA_CAKE_STATS_TIN_STATS,
+ __TCA_CAKE_STATS_MAX
+};
+#define TCA_CAKE_STATS_MAX (__TCA_CAKE_STATS_MAX - 1)
+
+enum {
+ __TCA_CAKE_TIN_STATS_INVALID,
+ TCA_CAKE_TIN_STATS_PAD,
+ TCA_CAKE_TIN_STATS_SENT_PACKETS,
+ TCA_CAKE_TIN_STATS_SENT_BYTES64,
+ TCA_CAKE_TIN_STATS_DROPPED_PACKETS,
+ TCA_CAKE_TIN_STATS_DROPPED_BYTES64,
+ TCA_CAKE_TIN_STATS_ACKS_DROPPED_PACKETS,
+ TCA_CAKE_TIN_STATS_ACKS_DROPPED_BYTES64,
+ TCA_CAKE_TIN_STATS_ECN_MARKED_PACKETS,
+ TCA_CAKE_TIN_STATS_ECN_MARKED_BYTES64,
+ TCA_CAKE_TIN_STATS_BACKLOG_PACKETS,
+ TCA_CAKE_TIN_STATS_BACKLOG_BYTES64,
+ TCA_CAKE_TIN_STATS_THRESHOLD_RATE64,
+ TCA_CAKE_TIN_STATS_TARGET_US,
+ TCA_CAKE_TIN_STATS_INTERVAL_US,
+ TCA_CAKE_TIN_STATS_WAY_INDIRECT_HITS,
+ TCA_CAKE_TIN_STATS_WAY_MISSES,
+ TCA_CAKE_TIN_STATS_WAY_COLLISIONS,
+ TCA_CAKE_TIN_STATS_PEAK_DELAY_US,
+ TCA_CAKE_TIN_STATS_AVG_DELAY_US,
+ TCA_CAKE_TIN_STATS_BASE_DELAY_US,
+ TCA_CAKE_TIN_STATS_SPARSE_FLOWS,
+ TCA_CAKE_TIN_STATS_BULK_FLOWS,
+ TCA_CAKE_TIN_STATS_UNRESPONSIVE_FLOWS,
+ TCA_CAKE_TIN_STATS_MAX_SKBLEN,
+ TCA_CAKE_TIN_STATS_FLOW_QUANTUM,
+ __TCA_CAKE_TIN_STATS_MAX
+};
+#define TCA_CAKE_TIN_STATS_MAX (__TCA_CAKE_TIN_STATS_MAX - 1)
+#define TC_CAKE_MAX_TINS (8)
+
+enum {
+ CAKE_FLOW_NONE = 0,
+ CAKE_FLOW_SRC_IP,
+ CAKE_FLOW_DST_IP,
+ CAKE_FLOW_HOSTS, /* = CAKE_FLOW_SRC_IP | CAKE_FLOW_DST_IP */
+ CAKE_FLOW_FLOWS,
+ CAKE_FLOW_DUAL_SRC, /* = CAKE_FLOW_SRC_IP | CAKE_FLOW_FLOWS */
+ CAKE_FLOW_DUAL_DST, /* = CAKE_FLOW_DST_IP | CAKE_FLOW_FLOWS */
+ CAKE_FLOW_TRIPLE, /* = CAKE_FLOW_HOSTS | CAKE_FLOW_FLOWS */
+ CAKE_FLOW_MAX,
+};
+
+enum {
+ CAKE_DIFFSERV_DIFFSERV3 = 0,
+ CAKE_DIFFSERV_DIFFSERV4,
+ CAKE_DIFFSERV_DIFFSERV8,
+ CAKE_DIFFSERV_BESTEFFORT,
+ CAKE_DIFFSERV_PRECEDENCE,
+ CAKE_DIFFSERV_MAX
+};
+
+enum {
+ CAKE_ACK_NONE = 0,
+ CAKE_ACK_FILTER,
+ CAKE_ACK_AGGRESSIVE,
+ CAKE_ACK_MAX
+};
+
+enum {
+ CAKE_ATM_NONE = 0,
+ CAKE_ATM_ATM,
+ CAKE_ATM_PTM,
+ CAKE_ATM_MAX
+};
+
#endif
diff --git a/net/sched/Kconfig b/net/sched/Kconfig
index a01169fb5325..6e7d614b5757 100644
--- a/net/sched/Kconfig
+++ b/net/sched/Kconfig
@@ -284,6 +284,17 @@ config NET_SCH_FQ_CODEL
If unsure, say N.
+config NET_SCH_CAKE
+ tristate "Common Applications Kept Enhanced (CAKE)"
+ help
+ Say Y here if you want to use the Common Applications Kept Enhanced
+ (CAKE) queue management algorithm.
+
+ To compile this driver as a module, choose M here: the module
+ will be called sch_cake.
+
+ If unsure, say N.
+
config NET_SCH_FQ
tristate "Fair Queue"
help
diff --git a/net/sched/Makefile b/net/sched/Makefile
index 8811d3804878..435054cee32c 100644
--- a/net/sched/Makefile
+++ b/net/sched/Makefile
@@ -50,6 +50,7 @@ obj-$(CONFIG_NET_SCH_CHOKE) += sch_choke.o
obj-$(CONFIG_NET_SCH_QFQ) += sch_qfq.o
obj-$(CONFIG_NET_SCH_CODEL) += sch_codel.o
obj-$(CONFIG_NET_SCH_FQ_CODEL) += sch_fq_codel.o
+obj-$(CONFIG_NET_SCH_CAKE) += sch_cake.o
obj-$(CONFIG_NET_SCH_FQ) += sch_fq.o
obj-$(CONFIG_NET_SCH_HHF) += sch_hhf.o
obj-$(CONFIG_NET_SCH_PIE) += sch_pie.o
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
new file mode 100644
index 000000000000..9d7fc1848778
--- /dev/null
+++ b/net/sched/sch_cake.c
@@ -0,0 +1,1730 @@
+// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause
+
+/* COMMON Applications Kept Enhanced (CAKE) discipline
+ *
+ * Copyright (C) 2014-2018 Jonathan Morton <chromatix99@gmail.com>
+ * Copyright (C) 2015-2018 Toke Høiland-Jørgensen <toke@toke.dk>
+ * Copyright (C) 2014-2018 Dave Täht <dave.taht@gmail.com>
+ * Copyright (C) 2015-2018 Sebastian Moeller <moeller0@gmx.de>
+ * (C) 2015-2018 Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk>
+ * Copyright (C) 2017 Ryan Mounce <ryan@mounce.com.au>
+ *
+ * The CAKE Principles:
+ * (or, how to have your cake and eat it too)
+ *
+ * This is a combination of several shaping, AQM and FQ techniques into one
+ * easy-to-use package:
+ *
+ * - An overall bandwidth shaper, to move the bottleneck away from dumb CPE
+ * equipment and bloated MACs. This operates in deficit mode (as in sch_fq),
+ * eliminating the need for any sort of burst parameter (eg. token bucket
+ * depth). Burst support is limited to that necessary to overcome scheduling
+ * latency.
+ *
+ * - A Diffserv-aware priority queue, giving more priority to certain classes,
+ * up to a specified fraction of bandwidth. Above that bandwidth threshold,
+ * the priority is reduced to avoid starving other tins.
+ *
+ * - Each priority tin has a separate Flow Queue system, to isolate traffic
+ * flows from each other. This prevents a burst on one flow from increasing
+ * the delay to another. Flows are distributed to queues using a
+ * set-associative hash function.
+ *
+ * - Each queue is actively managed by Cobalt, which is a combination of the
+ * Codel and Blue AQM algorithms. This serves flows fairly, and signals
+ * congestion early via ECN (if available) and/or packet drops, to keep
+ * latency low. The codel parameters are auto-tuned based on the bandwidth
+ * setting, as is necessary at low bandwidths.
+ *
+ * The configuration parameters are kept deliberately simple for ease of use.
+ * Everything has sane defaults. Complete generality of configuration is *not*
+ * a goal.
+ *
+ * The priority queue operates according to a weighted DRR scheme, combined with
+ * a bandwidth tracker which reuses the shaper logic to detect which side of the
+ * bandwidth sharing threshold the tin is operating. This determines whether a
+ * priority-based weight (high) or a bandwidth-based weight (low) is used for
+ * that tin in the current pass.
+ *
+ * This qdisc was inspired by Eric Dumazet's fq_codel code, which he kindly
+ * granted us permission to leverage.
+ */
+
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/jiffies.h>
+#include <linux/string.h>
+#include <linux/in.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/skbuff.h>
+#include <linux/jhash.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/reciprocal_div.h>
+#include <net/netlink.h>
+#include <linux/version.h>
+#include <linux/if_vlan.h>
+#include <net/pkt_sched.h>
+#include <net/pkt_cls.h>
+#include <net/tcp.h>
+#include <net/flow_dissector.h>
+
+#define CAKE_SET_WAYS (8)
+#define CAKE_MAX_TINS (8)
+#define CAKE_QUEUES (1024)
+#define CAKE_FLOW_MASK 63
+#define CAKE_FLOW_NAT_FLAG 64
+#define US2TIME(a) (a * (u64)NSEC_PER_USEC)
+
+typedef u64 cobalt_time_t;
+typedef s64 cobalt_tdiff_t;
+
+/* struct cobalt_params - contains codel and blue parameters
+ * @interval: codel initial drop rate
+ * @target: maximum persistent sojourn time & blue update rate
+ * @mtu_time: serialisation delay of maximum-size packet
+ * @p_inc: increment of blue drop probability (0.32 fxp)
+ * @p_dec: decrement of blue drop probability (0.32 fxp)
+ */
+struct cobalt_params {
+ cobalt_time_t interval;
+ cobalt_time_t target;
+ cobalt_time_t mtu_time;
+ u32 p_inc;
+ u32 p_dec;
+};
+
+/* struct cobalt_vars - contains codel and blue variables
+ * @count: codel dropping frequency
+ * @rec_inv_sqrt: reciprocal value of sqrt(count) >> 1
+ * @drop_next: time to drop next packet, or when we dropped last
+ * @blue_timer: Blue time to next drop
+ * @p_drop: BLUE drop probability (0.32 fxp)
+ * @dropping: set if in dropping state
+ * @ecn_marked: set if marked
+ */
+struct cobalt_vars {
+ u32 count;
+ u32 rec_inv_sqrt;
+ cobalt_time_t drop_next;
+ cobalt_time_t blue_timer;
+ u32 p_drop;
+ bool dropping;
+ bool ecn_marked;
+};
+
+enum {
+ CAKE_SET_NONE = 0,
+ CAKE_SET_SPARSE,
+ CAKE_SET_SPARSE_WAIT, /* counted in SPARSE, actually in BULK */
+ CAKE_SET_BULK,
+ CAKE_SET_DECAYING
+};
+
+struct cake_flow {
+ /* this stuff is all needed per-flow at dequeue time */
+ struct sk_buff *head;
+ struct sk_buff *tail;
+ struct list_head flowchain;
+ s32 deficit;
+ struct cobalt_vars cvars;
+ u16 srchost; /* index into cake_host table */
+ u16 dsthost;
+ u8 set;
+}; /* please try to keep this structure <= 64 bytes */
+
+struct cake_host {
+ u32 srchost_tag;
+ u32 dsthost_tag;
+ u16 srchost_refcnt;
+ u16 dsthost_refcnt;
+};
+
+struct cake_heap_entry {
+ u16 t:3, b:10;
+};
+
+struct cake_tin_data {
+ struct cake_flow flows[CAKE_QUEUES];
+ u32 backlogs[CAKE_QUEUES];
+ u32 tags[CAKE_QUEUES]; /* for set association */
+ u16 overflow_idx[CAKE_QUEUES];
+ struct cake_host hosts[CAKE_QUEUES]; /* for triple isolation */
+ u16 flow_quantum;
+
+ struct cobalt_params cparams;
+ u32 drop_overlimit;
+ u16 bulk_flow_count;
+ u16 sparse_flow_count;
+ u16 decaying_flow_count;
+ u16 unresponsive_flow_count;
+
+ u32 max_skblen;
+
+ struct list_head new_flows;
+ struct list_head old_flows;
+ struct list_head decaying_flows;
+
+ /* time_next = time_this + ((len * rate_ns) >> rate_shft) */
+ u64 tin_time_next_packet;
+ u64 tin_rate_ns;
+ u64 tin_rate_bps;
+ u16 tin_rate_shft;
+
+ u16 tin_quantum_prio;
+ u16 tin_quantum_band;
+ s32 tin_deficit;
+ u32 tin_backlog;
+ u32 tin_dropped;
+ u32 tin_ecn_mark;
+
+ u32 packets;
+ u64 bytes;
+
+ u32 ack_drops;
+
+ /* moving averages */
+ cobalt_time_t avge_delay;
+ cobalt_time_t peak_delay;
+ cobalt_time_t base_delay;
+
+ /* hash function stats */
+ u32 way_directs;
+ u32 way_hits;
+ u32 way_misses;
+ u32 way_collisions;
+}; /* number of tins is small, so size of this struct doesn't matter much */
+
+struct cake_sched_data {
+ struct tcf_proto __rcu *filter_list; /* optional external classifier */
+ struct tcf_block *block;
+ struct cake_tin_data *tins;
+
+ struct cake_heap_entry overflow_heap[CAKE_QUEUES * CAKE_MAX_TINS];
+ u16 overflow_timeout;
+
+ u16 tin_cnt;
+ u8 tin_mode;
+ u8 flow_mode;
+ u8 ack_filter;
+ u8 atm_mode;
+
+ /* time_next = time_this + ((len * rate_ns) >> rate_shft) */
+ u16 rate_shft;
+ u64 time_next_packet;
+ u64 failsafe_next_packet;
+ u64 rate_ns;
+ u64 rate_bps;
+ u16 rate_flags;
+ s16 rate_overhead;
+ u16 rate_mpu;
+ u32 interval;
+ u32 target;
+
+ /* resource tracking */
+ u32 buffer_used;
+ u32 buffer_max_used;
+ u32 buffer_limit;
+ u32 buffer_config_limit;
+
+ /* indices for dequeue */
+ u16 cur_tin;
+ u16 cur_flow;
+
+ struct qdisc_watchdog watchdog;
+ const u8 *tin_index;
+ const u8 *tin_order;
+
+ /* bandwidth capacity estimate */
+ u64 last_packet_time;
+ u64 avg_packet_interval;
+ u64 avg_window_begin;
+ u64 avg_window_bytes;
+ u64 avg_peak_bandwidth;
+ u64 last_reconfig_time;
+
+ /* packet length stats */
+ u32 avg_netoff;
+ u16 max_netlen;
+ u16 max_adjlen;
+ u16 min_netlen;
+ u16 min_adjlen;
+};
+
+enum {
+ CAKE_FLAG_OVERHEAD = BIT(0),
+ CAKE_FLAG_AUTORATE_INGRESS = BIT(1),
+ CAKE_FLAG_INGRESS = BIT(2),
+ CAKE_FLAG_WASH = BIT(3),
+ CAKE_FLAG_SPLIT_GSO = BIT(4)
+};
+
+/* COBALT operates the Codel and BLUE algorithms in parallel, in order to
+ * obtain the best features of each. Codel is excellent on flows which
+ * respond to congestion signals in a TCP-like way. BLUE is more effective on
+ * unresponsive flows.
+ */
+
+struct cobalt_skb_cb {
+ cobalt_time_t enqueue_time;
+};
+
+static cobalt_time_t cobalt_get_time(void)
+{
+ return ktime_get_ns();
+}
+
+static u32 cobalt_time_to_us(cobalt_time_t val)
+{
+ do_div(val, NSEC_PER_USEC);
+ return (u32)val;
+}
+
+static struct cobalt_skb_cb *get_cobalt_cb(const struct sk_buff *skb)
+{
+ qdisc_cb_private_validate(skb, sizeof(struct cobalt_skb_cb));
+ return (struct cobalt_skb_cb *)qdisc_skb_cb(skb)->data;
+}
+
+static cobalt_time_t cobalt_get_enqueue_time(const struct sk_buff *skb)
+{
+ return get_cobalt_cb(skb)->enqueue_time;
+}
+
+static void cobalt_set_enqueue_time(struct sk_buff *skb,
+ cobalt_time_t now)
+{
+ get_cobalt_cb(skb)->enqueue_time = now;
+}
+
+static u16 quantum_div[CAKE_QUEUES + 1] = {0};
+
+#define REC_INV_SQRT_CACHE (16)
+static u32 cobalt_rec_inv_sqrt_cache[REC_INV_SQRT_CACHE] = {0};
+
+/* http://en.wikipedia.org/wiki/Methods_of_computing_square_roots
+ * new_invsqrt = (invsqrt / 2) * (3 - count * invsqrt^2)
+ *
+ * Here, invsqrt is a fixed point number (< 1.0), 32bit mantissa, aka Q0.32
+ */
+
+static void cobalt_newton_step(struct cobalt_vars *vars)
+{
+ u32 invsqrt, invsqrt2;
+ u64 val;
+
+ invsqrt = vars->rec_inv_sqrt;
+ invsqrt2 = ((u64)invsqrt * invsqrt) >> 32;
+ val = (3LL << 32) - ((u64)vars->count * invsqrt2);
+
+ val >>= 2; /* avoid overflow in following multiply */
+ val = (val * invsqrt) >> (32 - 2 + 1);
+
+ vars->rec_inv_sqrt = val;
+}
+
+static void cobalt_invsqrt(struct cobalt_vars *vars)
+{
+ if (vars->count < REC_INV_SQRT_CACHE)
+ vars->rec_inv_sqrt = cobalt_rec_inv_sqrt_cache[vars->count];
+ else
+ cobalt_newton_step(vars);
+}
+
+/* There is a big difference in timing between the accurate values placed in
+ * the cache and the approximations given by a single Newton step for small
+ * count values, particularly when stepping from count 1 to 2 or vice versa.
+ * Above 16, a single Newton step gives sufficient accuracy in either
+ * direction, given the precision stored.
+ *
+ * The magnitude of the error when stepping up to count 2 is such as to give
+ * the value that *should* have been produced at count 4.
+ */
+
+static void cobalt_cache_init(void)
+{
+ struct cobalt_vars v;
+
+ memset(&v, 0, sizeof(v));
+ v.rec_inv_sqrt = ~0U;
+ cobalt_rec_inv_sqrt_cache[0] = v.rec_inv_sqrt;
+
+ for (v.count = 1; v.count < REC_INV_SQRT_CACHE; v.count++) {
+ cobalt_newton_step(&v);
+ cobalt_newton_step(&v);
+ cobalt_newton_step(&v);
+ cobalt_newton_step(&v);
+
+ cobalt_rec_inv_sqrt_cache[v.count] = v.rec_inv_sqrt;
+ }
+}
+
+static void cobalt_vars_init(struct cobalt_vars *vars)
+{
+ memset(vars, 0, sizeof(*vars));
+
+ if (!cobalt_rec_inv_sqrt_cache[0]) {
+ cobalt_cache_init();
+ cobalt_rec_inv_sqrt_cache[0] = ~0;
+ }
+}
+
+/* CoDel control_law is t + interval/sqrt(count)
+ * We maintain in rec_inv_sqrt the reciprocal value of sqrt(count) to avoid
+ * both sqrt() and divide operation.
+ */
+static cobalt_time_t cobalt_control(cobalt_time_t t,
+ cobalt_time_t interval,
+ u32 rec_inv_sqrt)
+{
+ return t + reciprocal_scale(interval, rec_inv_sqrt);
+}
+
+/* Call this when a packet had to be dropped due to queue overflow. Returns
+ * true if the BLUE state was quiescent before but active after this call.
+ */
+static bool cobalt_queue_full(struct cobalt_vars *vars,
+ struct cobalt_params *p,
+ cobalt_time_t now)
+{
+ bool up = false;
+
+ if ((now - vars->blue_timer) > p->target) {
+ up = !vars->p_drop;
+ vars->p_drop += p->p_inc;
+ if (vars->p_drop < p->p_inc)
+ vars->p_drop = ~0;
+ vars->blue_timer = now;
+ }
+ vars->dropping = true;
+ vars->drop_next = now;
+ if (!vars->count)
+ vars->count = 1;
+
+ return up;
+}
+
+/* Call this when the queue was serviced but turned out to be empty. Returns
+ * true if the BLUE state was active before but quiescent after this call.
+ */
+static bool cobalt_queue_empty(struct cobalt_vars *vars,
+ struct cobalt_params *p,
+ cobalt_time_t now)
+{
+ bool down = false;
+
+ if (vars->p_drop && (now - vars->blue_timer) > p->target) {
+ if (vars->p_drop < p->p_dec)
+ vars->p_drop = 0;
+ else
+ vars->p_drop -= p->p_dec;
+ vars->blue_timer = now;
+ down = !vars->p_drop;
+ }
+ vars->dropping = false;
+
+ if (vars->count && (now - vars->drop_next) >= 0) {
+ vars->count--;
+ cobalt_invsqrt(vars);
+ vars->drop_next = cobalt_control(vars->drop_next,
+ p->interval,
+ vars->rec_inv_sqrt);
+ }
+
+ return down;
+}
+
+/* Call this with a freshly dequeued packet for possible congestion marking.
+ * Returns true as an instruction to drop the packet, false for delivery.
+ */
+static bool cobalt_should_drop(struct cobalt_vars *vars,
+ struct cobalt_params *p,
+ cobalt_time_t now,
+ struct sk_buff *skb)
+{
+ bool next_due, over_target, drop = false;
+ cobalt_tdiff_t sojourn, schedule;
+
+/* The 'schedule' variable records, in its sign, whether 'now' is before or
+ * after 'drop_next'. This allows 'drop_next' to be updated before the next
+ * scheduling decision is actually branched, without destroying that
+ * information. Similarly, the first 'schedule' value calculated is preserved
+ * in the boolean 'next_due'.
+ *
+ * As for 'drop_next', we take advantage of the fact that 'interval' is both
+ * the delay between first exceeding 'target' and the first signalling event,
+ * *and* the scaling factor for the signalling frequency. It's therefore very
+ * natural to use a single mechanism for both purposes, and eliminates a
+ * significant amount of reference Codel's spaghetti code. To help with this,
+ * both the '0' and '1' entries in the invsqrt cache are 0xFFFFFFFF, as close
+ * as possible to 1.0 in fixed-point.
+ */
+
+ sojourn = now - cobalt_get_enqueue_time(skb);
+ schedule = now - vars->drop_next;
+ over_target = sojourn > p->target &&
+ sojourn > p->mtu_time * 4;
+ next_due = vars->count && schedule >= 0;
+
+ vars->ecn_marked = false;
+
+ if (over_target) {
+ if (!vars->dropping) {
+ vars->dropping = true;
+ vars->drop_next = cobalt_control(now,
+ p->interval,
+ vars->rec_inv_sqrt);
+ }
+ if (!vars->count)
+ vars->count = 1;
+ } else if (vars->dropping) {
+ vars->dropping = false;
+ }
+
+ if (next_due && vars->dropping) {
+ /* Use ECN mark if possible, otherwise drop */
+ drop = !(vars->ecn_marked = INET_ECN_set_ce(skb));
+
+ vars->count++;
+ if (!vars->count)
+ vars->count--;
+ cobalt_invsqrt(vars);
+ vars->drop_next = cobalt_control(vars->drop_next,
+ p->interval,
+ vars->rec_inv_sqrt);
+ schedule = now - vars->drop_next;
+ } else {
+ while (next_due) {
+ vars->count--;
+ cobalt_invsqrt(vars);
+ vars->drop_next = cobalt_control(vars->drop_next,
+ p->interval,
+ vars->rec_inv_sqrt);
+ schedule = now - vars->drop_next;
+ next_due = vars->count && schedule >= 0;
+ }
+ }
+
+ /* Simple BLUE implementation. Lack of ECN is deliberate. */
+ if (vars->p_drop)
+ drop |= (prandom_u32() < vars->p_drop);
+
+ /* Overload the drop_next field as an activity timeout */
+ if (!vars->count)
+ vars->drop_next = now + p->interval;
+ else if (schedule > 0 && !drop)
+ vars->drop_next = now;
+
+ return drop;
+}
+
+/* Cake has several subtle multiple bit settings. In these cases you
+ * would be matching triple isolate mode as well.
+ */
+
+static bool cake_dsrc(int flow_mode)
+{
+ return (flow_mode & CAKE_FLOW_DUAL_SRC) == CAKE_FLOW_DUAL_SRC;
+}
+
+static bool cake_ddst(int flow_mode)
+{
+ return (flow_mode & CAKE_FLOW_DUAL_DST) == CAKE_FLOW_DUAL_DST;
+}
+
+static u32 cake_hash(struct cake_tin_data *q, const struct sk_buff *skb,
+ int flow_mode)
+{
+ u32 flow_hash = 0, srchost_hash, dsthost_hash;
+ u16 reduced_hash, srchost_idx, dsthost_idx;
+ struct flow_keys keys, host_keys;
+
+ if (unlikely(flow_mode == CAKE_FLOW_NONE))
+ return 0;
+
+ skb_flow_dissect_flow_keys(skb, &keys,
+ FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
+
+ /* flow_hash_from_keys() sorts the addresses by value, so we have
+ * to preserve their order in a separate data structure to treat
+ * src and dst host addresses as independently selectable.
+ */
+ host_keys = keys;
+ host_keys.ports.ports = 0;
+ host_keys.basic.ip_proto = 0;
+ host_keys.keyid.keyid = 0;
+ host_keys.tags.flow_label = 0;
+
+ switch (host_keys.control.addr_type) {
+ case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
+ host_keys.addrs.v4addrs.src = 0;
+ dsthost_hash = flow_hash_from_keys(&host_keys);
+ host_keys.addrs.v4addrs.src = keys.addrs.v4addrs.src;
+ host_keys.addrs.v4addrs.dst = 0;
+ srchost_hash = flow_hash_from_keys(&host_keys);
+ break;
+
+ case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
+ memset(&host_keys.addrs.v6addrs.src, 0,
+ sizeof(host_keys.addrs.v6addrs.src));
+ dsthost_hash = flow_hash_from_keys(&host_keys);
+ host_keys.addrs.v6addrs.src = keys.addrs.v6addrs.src;
+ memset(&host_keys.addrs.v6addrs.dst, 0,
+ sizeof(host_keys.addrs.v6addrs.dst));
+ srchost_hash = flow_hash_from_keys(&host_keys);
+ break;
+
+ default:
+ dsthost_hash = 0;
+ srchost_hash = 0;
+ }
+
+ /* This *must* be after the above switch, since as a
+ * side-effect it sorts the src and dst addresses.
+ */
+ if (flow_mode & CAKE_FLOW_FLOWS)
+ flow_hash = flow_hash_from_keys(&keys);
+
+ if (!(flow_mode & CAKE_FLOW_FLOWS)) {
+ if (flow_mode & CAKE_FLOW_SRC_IP)
+ flow_hash ^= srchost_hash;
+
+ if (flow_mode & CAKE_FLOW_DST_IP)
+ flow_hash ^= dsthost_hash;
+ }
+
+ reduced_hash = flow_hash % CAKE_QUEUES;
+
+ /* set-associative hashing */
+ /* fast path if no hash collision (direct lookup succeeds) */
+ if (likely(q->tags[reduced_hash] == flow_hash &&
+ q->flows[reduced_hash].set)) {
+ q->way_directs++;
+ } else {
+ u32 inner_hash = reduced_hash % CAKE_SET_WAYS;
+ u32 outer_hash = reduced_hash - inner_hash;
+ bool allocate_src = false;
+ bool allocate_dst = false;
+ u32 i, k;
+
+ /* check if any active queue in the set is reserved for
+ * this flow.
+ */
+ for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
+ i++, k = (k + 1) % CAKE_SET_WAYS) {
+ if (q->tags[outer_hash + k] == flow_hash) {
+ if (i)
+ q->way_hits++;
+
+ if (!q->flows[outer_hash + k].set) {
+ /* need to increment host refcnts */
+ allocate_src = cake_dsrc(flow_mode);
+ allocate_dst = cake_ddst(flow_mode);
+ }
+
+ goto found;
+ }
+ }
+
+ /* no queue is reserved for this flow, look for an
+ * empty one.
+ */
+ for (i = 0; i < CAKE_SET_WAYS;
+ i++, k = (k + 1) % CAKE_SET_WAYS) {
+ if (!q->flows[outer_hash + k].set) {
+ q->way_misses++;
+ allocate_src = cake_dsrc(flow_mode);
+ allocate_dst = cake_ddst(flow_mode);
+ goto found;
+ }
+ }
+
+ /* With no empty queues, default to the original
+ * queue, accept the collision, update the host tags.
+ */
+ q->way_collisions++;
+ q->hosts[q->flows[reduced_hash].srchost].srchost_refcnt--;
+ q->hosts[q->flows[reduced_hash].dsthost].dsthost_refcnt--;
+ allocate_src = cake_dsrc(flow_mode);
+ allocate_dst = cake_ddst(flow_mode);
+found:
+ /* reserve queue for future packets in same flow */
+ reduced_hash = outer_hash + k;
+ q->tags[reduced_hash] = flow_hash;
+
+ if (allocate_src) {
+ srchost_idx = srchost_hash % CAKE_QUEUES;
+ inner_hash = srchost_idx % CAKE_SET_WAYS;
+ outer_hash = srchost_idx - inner_hash;
+ for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
+ i++, k = (k + 1) % CAKE_SET_WAYS) {
+ if (q->hosts[outer_hash + k].srchost_tag ==
+ srchost_hash)
+ goto found_src;
+ }
+ for (i = 0; i < CAKE_SET_WAYS;
+ i++, k = (k + 1) % CAKE_SET_WAYS) {
+ if (!q->hosts[outer_hash + k].srchost_refcnt)
+ break;
+ }
+ q->hosts[outer_hash + k].srchost_tag = srchost_hash;
+found_src:
+ srchost_idx = outer_hash + k;
+ q->hosts[srchost_idx].srchost_refcnt++;
+ q->flows[reduced_hash].srchost = srchost_idx;
+ }
+
+ if (allocate_dst) {
+ dsthost_idx = dsthost_hash % CAKE_QUEUES;
+ inner_hash = dsthost_idx % CAKE_SET_WAYS;
+ outer_hash = dsthost_idx - inner_hash;
+ for (i = 0, k = inner_hash; i < CAKE_SET_WAYS;
+ i++, k = (k + 1) % CAKE_SET_WAYS) {
+ if (q->hosts[outer_hash + k].dsthost_tag ==
+ dsthost_hash)
+ goto found_dst;
+ }
+ for (i = 0; i < CAKE_SET_WAYS;
+ i++, k = (k + 1) % CAKE_SET_WAYS) {
+ if (!q->hosts[outer_hash + k].dsthost_refcnt)
+ break;
+ }
+ q->hosts[outer_hash + k].dsthost_tag = dsthost_hash;
+found_dst:
+ dsthost_idx = outer_hash + k;
+ q->hosts[dsthost_idx].dsthost_refcnt++;
+ q->flows[reduced_hash].dsthost = dsthost_idx;
+ }
+ }
+
+ return reduced_hash;
+}
+
+static u32 cake_classify(struct Qdisc *sch, struct cake_tin_data *t,
+ struct sk_buff *skb, int flow_mode, int *qerr)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct tcf_proto *filter;
+ struct tcf_result res;
+ int result;
+
+ filter = rcu_dereference_bh(q->filter_list);
+ if (!filter)
+ return cake_hash(t, skb, flow_mode) + 1;
+
+ *qerr = NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
+ result = tcf_classify(skb, filter, &res, false);
+ if (result >= 0) {
+#ifdef CONFIG_NET_CLS_ACT
+ switch (result) {
+ case TC_ACT_STOLEN:
+ case TC_ACT_QUEUED:
+ case TC_ACT_TRAP:
+ *qerr = NET_XMIT_SUCCESS | __NET_XMIT_STOLEN;
+ /* fall through */
+ case TC_ACT_SHOT:
+ return 0;
+ }
+#endif
+ if (TC_H_MIN(res.classid) <= CAKE_QUEUES)
+ return TC_H_MIN(res.classid);
+ }
+ return 0;
+}
+
+/* helper functions : might be changed when/if skb use a standard list_head */
+/* remove one skb from head of slot queue */
+
+static struct sk_buff *dequeue_head(struct cake_flow *flow)
+{
+ struct sk_buff *skb = flow->head;
+
+ if (skb) {
+ flow->head = skb->next;
+ skb->next = NULL;
+ }
+
+ return skb;
+}
+
+/* add skb to flow queue (tail add) */
+
+static void flow_queue_add(struct cake_flow *flow, struct sk_buff *skb)
+{
+ if (!flow->head)
+ flow->head = skb;
+ else
+ flow->tail->next = skb;
+ flow->tail = skb;
+ skb->next = NULL;
+}
+
+static cobalt_time_t cake_ewma(cobalt_time_t avg, cobalt_time_t sample,
+ u32 shift)
+{
+ avg -= avg >> shift;
+ avg += sample >> shift;
+ return avg;
+}
+
+static void cake_heap_swap(struct cake_sched_data *q, u16 i, u16 j)
+{
+ struct cake_heap_entry ii = q->overflow_heap[i];
+ struct cake_heap_entry jj = q->overflow_heap[j];
+
+ q->overflow_heap[i] = jj;
+ q->overflow_heap[j] = ii;
+
+ q->tins[ii.t].overflow_idx[ii.b] = j;
+ q->tins[jj.t].overflow_idx[jj.b] = i;
+}
+
+static u32 cake_heap_get_backlog(const struct cake_sched_data *q, u16 i)
+{
+ struct cake_heap_entry ii = q->overflow_heap[i];
+
+ return q->tins[ii.t].backlogs[ii.b];
+}
+
+static void cake_heapify(struct cake_sched_data *q, u16 i)
+{
+ static const u32 a = CAKE_MAX_TINS * CAKE_QUEUES;
+ u32 mb = cake_heap_get_backlog(q, i);
+ u32 m = i;
+
+ while (m < a) {
+ u32 l = m + m + 1;
+ u32 r = l + 1;
+
+ if (l < a) {
+ u32 lb = cake_heap_get_backlog(q, l);
+
+ if (lb > mb) {
+ m = l;
+ mb = lb;
+ }
+ }
+
+ if (r < a) {
+ u32 rb = cake_heap_get_backlog(q, r);
+
+ if (rb > mb) {
+ m = r;
+ mb = rb;
+ }
+ }
+
+ if (m != i) {
+ cake_heap_swap(q, i, m);
+ i = m;
+ } else {
+ break;
+ }
+ }
+}
+
+static void cake_heapify_up(struct cake_sched_data *q, u16 i)
+{
+ while (i > 0 && i < CAKE_MAX_TINS * CAKE_QUEUES) {
+ u16 p = (i - 1) >> 1;
+ u32 ib = cake_heap_get_backlog(q, i);
+ u32 pb = cake_heap_get_backlog(q, p);
+
+ if (ib > pb) {
+ cake_heap_swap(q, i, p);
+ i = p;
+ } else {
+ break;
+ }
+ }
+}
+
+static int cake_advance_shaper(struct cake_sched_data *q,
+ struct cake_tin_data *b,
+ struct sk_buff *skb,
+ u64 now, bool drop)
+{
+ u32 len = qdisc_pkt_len(skb);
+
+ /* charge packet bandwidth to this tin
+ * and to the global shaper.
+ */
+ if (q->rate_ns) {
+ s64 tdiff1 = b->tin_time_next_packet - now;
+ s64 tdiff2 = (len * b->tin_rate_ns) >> b->tin_rate_shft;
+ s64 tdiff3 = (len * q->rate_ns) >> q->rate_shft;
+ s64 tdiff4 = tdiff3 + (tdiff3 >> 1);
+
+ if (tdiff1 < 0)
+ b->tin_time_next_packet += tdiff2;
+ else if (tdiff1 < tdiff2)
+ b->tin_time_next_packet = now + tdiff2;
+
+ q->time_next_packet += tdiff3;
+ if (!drop)
+ q->failsafe_next_packet += tdiff4;
+ }
+ return len;
+}
+
+static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ u64 now = cobalt_get_time();
+ u32 idx = 0, tin = 0, len;
+ struct cake_heap_entry qq;
+ struct cake_tin_data *b;
+ struct cake_flow *flow;
+ struct sk_buff *skb;
+
+ if (!q->overflow_timeout) {
+ int i;
+ /* Build fresh max-heap */
+ for (i = CAKE_MAX_TINS * CAKE_QUEUES / 2; i >= 0; i--)
+ cake_heapify(q, i);
+ }
+ q->overflow_timeout = 65535;
+
+ /* select longest queue for pruning */
+ qq = q->overflow_heap[0];
+ tin = qq.t;
+ idx = qq.b;
+
+ b = &q->tins[tin];
+ flow = &b->flows[idx];
+ skb = dequeue_head(flow);
+ if (unlikely(!skb)) {
+ /* heap has gone wrong, rebuild it next time */
+ q->overflow_timeout = 0;
+ return idx + (tin << 16);
+ }
+
+ if (cobalt_queue_full(&flow->cvars, &b->cparams, now))
+ b->unresponsive_flow_count++;
+
+ len = qdisc_pkt_len(skb);
+ q->buffer_used -= skb->truesize;
+ b->backlogs[idx] -= len;
+ b->tin_backlog -= len;
+ sch->qstats.backlog -= len;
+ qdisc_tree_reduce_backlog(sch, 1, len);
+
+ b->tin_dropped++;
+ sch->qstats.drops++;
+
+ __qdisc_drop(skb, to_free);
+ sch->q.qlen--;
+
+ cake_heapify(q, 0);
+
+ return idx + (tin << 16);
+}
+
+static void cake_reconfigure(struct Qdisc *sch);
+
+static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
+ struct sk_buff **to_free)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ int len = qdisc_pkt_len(skb);
+ u64 now = cobalt_get_time();
+ int uninitialized_var(ret);
+ struct cake_tin_data *b;
+ struct cake_flow *flow;
+ u32 idx, tin;
+
+ tin = 0;
+ b = &q->tins[tin];
+
+ /* choose flow to insert into */
+ idx = cake_classify(sch, b, skb, q->flow_mode, &ret);
+ if (idx == 0) {
+ if (ret & __NET_XMIT_BYPASS)
+ qdisc_qstats_drop(sch);
+ __qdisc_drop(skb, to_free);
+ return ret;
+ }
+ idx--;
+ flow = &b->flows[idx];
+
+ /* ensure shaper state isn't stale */
+ if (!b->tin_backlog) {
+ if (b->tin_time_next_packet < now)
+ b->tin_time_next_packet = now;
+
+ if (!sch->q.qlen) {
+ if (q->time_next_packet < now) {
+ q->failsafe_next_packet = now;
+ q->time_next_packet = now;
+ } else if (q->time_next_packet > now &&
+ q->failsafe_next_packet > now) {
+ u64 next = min(q->time_next_packet,
+ q->failsafe_next_packet);
+ sch->qstats.overlimits++;
+ qdisc_watchdog_schedule_ns(&q->watchdog, next);
+ }
+ }
+ }
+
+ if (unlikely(len > b->max_skblen))
+ b->max_skblen = len;
+
+ cobalt_set_enqueue_time(skb, now);
+ flow_queue_add(flow, skb);
+
+ sch->q.qlen++;
+ q->buffer_used += skb->truesize;
+
+ /* stats */
+ b->packets++;
+ b->bytes += len;
+ b->backlogs[idx] += len;
+ b->tin_backlog += len;
+ sch->qstats.backlog += len;
+ q->avg_window_bytes += len;
+
+ if (q->overflow_timeout)
+ cake_heapify_up(q, b->overflow_idx[idx]);
+
+ /* incoming bandwidth capacity estimate */
+ q->avg_window_bytes = 0;
+ q->last_packet_time = now;
+
+ /* flowchain */
+ if (!flow->set || flow->set == CAKE_SET_DECAYING) {
+ struct cake_host *srchost = &b->hosts[flow->srchost];
+ struct cake_host *dsthost = &b->hosts[flow->dsthost];
+ u16 host_load = 1;
+
+ if (!flow->set) {
+ list_add_tail(&flow->flowchain, &b->new_flows);
+ } else {
+ b->decaying_flow_count--;
+ list_move_tail(&flow->flowchain, &b->new_flows);
+ }
+ flow->set = CAKE_SET_SPARSE;
+ b->sparse_flow_count++;
+
+ if (cake_dsrc(q->flow_mode))
+ host_load = max(host_load, srchost->srchost_refcnt);
+
+ if (cake_ddst(q->flow_mode))
+ host_load = max(host_load, dsthost->dsthost_refcnt);
+
+ flow->deficit = (b->flow_quantum *
+ quantum_div[host_load]) >> 16;
+ } else if (flow->set == CAKE_SET_SPARSE_WAIT) {
+ /* this flow was empty, accounted as a sparse flow, but actually
+ * in the bulk rotation.
+ */
+ flow->set = CAKE_SET_BULK;
+ b->sparse_flow_count--;
+ b->bulk_flow_count++;
+ }
+
+ if (q->buffer_used > q->buffer_max_used)
+ q->buffer_max_used = q->buffer_used;
+
+ if (q->buffer_used > q->buffer_limit) {
+ u32 dropped = 0;
+
+ while (q->buffer_used > q->buffer_limit) {
+ dropped++;
+ cake_drop(sch, to_free);
+ }
+ b->drop_overlimit += dropped;
+ }
+ return NET_XMIT_SUCCESS;
+}
+
+static struct sk_buff *cake_dequeue_one(struct Qdisc *sch)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct cake_tin_data *b = &q->tins[q->cur_tin];
+ struct cake_flow *flow = &b->flows[q->cur_flow];
+ struct sk_buff *skb = NULL;
+ u32 len;
+
+ if (flow->head) {
+ skb = dequeue_head(flow);
+ len = qdisc_pkt_len(skb);
+ b->backlogs[q->cur_flow] -= len;
+ b->tin_backlog -= len;
+ sch->qstats.backlog -= len;
+ q->buffer_used -= skb->truesize;
+ sch->q.qlen--;
+
+ if (q->overflow_timeout)
+ cake_heapify(q, b->overflow_idx[q->cur_flow]);
+ }
+ return skb;
+}
+
+/* Discard leftover packets from a tin no longer in use. */
+static void cake_clear_tin(struct Qdisc *sch, u16 tin)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct sk_buff *skb;
+
+ q->cur_tin = tin;
+ for (q->cur_flow = 0; q->cur_flow < CAKE_QUEUES; q->cur_flow++)
+ while (!!(skb = cake_dequeue_one(sch)))
+ kfree_skb(skb);
+}
+
+static struct sk_buff *cake_dequeue(struct Qdisc *sch)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct cake_tin_data *b = &q->tins[q->cur_tin];
+ struct cake_host *srchost, *dsthost;
+ cobalt_time_t now = ktime_get_ns();
+ struct cake_flow *flow;
+ struct list_head *head;
+ bool first_flow = true;
+ cobalt_time_t delay;
+ struct sk_buff *skb;
+ u16 host_load;
+ u32 len;
+
+begin:
+ if (!sch->q.qlen)
+ return NULL;
+
+ /* global hard shaper */
+ if (q->time_next_packet > now && q->failsafe_next_packet > now) {
+ u64 next = min(q->time_next_packet, q->failsafe_next_packet);
+
+ sch->qstats.overlimits++;
+ qdisc_watchdog_schedule_ns(&q->watchdog, next);
+ return NULL;
+ }
+
+ /* Choose a class to work on. */
+ if (!q->rate_ns) {
+ /* In unlimited mode, can't rely on shaper timings, just balance
+ * with DRR
+ */
+ while (b->tin_deficit < 0 ||
+ !(b->sparse_flow_count + b->bulk_flow_count)) {
+ if (b->tin_deficit <= 0)
+ b->tin_deficit += b->tin_quantum_band;
+
+ q->cur_tin++;
+ b++;
+ if (q->cur_tin >= q->tin_cnt) {
+ q->cur_tin = 0;
+ b = q->tins;
+ }
+ }
+ } else {
+ /* In shaped mode, choose:
+ * - Highest-priority tin with queue and meeting schedule, or
+ * - The earliest-scheduled tin with queue.
+ */
+ s64 best_time = 0xFFFFFFFFFFFFUL;
+ int tin, best_tin = 0;
+
+ for (tin = 0; tin < q->tin_cnt; tin++) {
+ b = q->tins + tin;
+ if ((b->sparse_flow_count + b->bulk_flow_count) > 0) {
+ s64 tdiff = b->tin_time_next_packet - now;
+
+ if (tdiff <= 0 || tdiff <= best_time) {
+ best_time = tdiff;
+ best_tin = tin;
+ }
+ }
+ }
+
+ q->cur_tin = best_tin;
+ b = q->tins + best_tin;
+ }
+
+retry:
+ /* service this class */
+ head = &b->decaying_flows;
+ if (!first_flow || list_empty(head)) {
+ head = &b->new_flows;
+ if (list_empty(head)) {
+ head = &b->old_flows;
+ if (unlikely(list_empty(head))) {
+ head = &b->decaying_flows;
+ if (unlikely(list_empty(head)))
+ goto begin;
+ }
+ }
+ }
+ flow = list_first_entry(head, struct cake_flow, flowchain);
+ q->cur_flow = flow - b->flows;
+ first_flow = false;
+
+ /* triple isolation (modified DRR++) */
+ srchost = &b->hosts[flow->srchost];
+ dsthost = &b->hosts[flow->dsthost];
+ host_load = 1;
+
+ if (cake_dsrc(q->flow_mode))
+ host_load = max(host_load, srchost->srchost_refcnt);
+
+ if (cake_ddst(q->flow_mode))
+ host_load = max(host_load, dsthost->dsthost_refcnt);
+
+ WARN_ON(host_load > CAKE_QUEUES);
+
+ /* flow isolation (DRR++) */
+ if (flow->deficit <= 0) {
+ /* The shifted prandom_u32() is a way to apply dithering to
+ * avoid accumulating roundoff errors
+ */
+ flow->deficit += (b->flow_quantum * quantum_div[host_load] +
+ (prandom_u32() >> 16)) >> 16;
+ list_move_tail(&flow->flowchain, &b->old_flows);
+
+ /* Keep all flows with deficits out of the sparse and decaying
+ * rotations. No non-empty flow can go into the decaying
+ * rotation, so they can't get deficits
+ */
+ if (flow->set == CAKE_SET_SPARSE) {
+ if (flow->head) {
+ b->sparse_flow_count--;
+ b->bulk_flow_count++;
+ flow->set = CAKE_SET_BULK;
+ } else {
+ /* we've moved it to the bulk rotation for
+ * correct deficit accounting but we still want
+ * to count it as a sparse flow, not a bulk one.
+ */
+ flow->set = CAKE_SET_SPARSE_WAIT;
+ }
+ }
+ goto retry;
+ }
+
+ /* Retrieve a packet via the AQM */
+ while (1) {
+ skb = cake_dequeue_one(sch);
+ if (!skb) {
+ /* this queue was actually empty */
+ if (cobalt_queue_empty(&flow->cvars, &b->cparams, now))
+ b->unresponsive_flow_count--;
+
+ if (flow->cvars.p_drop || flow->cvars.count ||
+ now < flow->cvars.drop_next) {
+ /* keep in the flowchain until the state has
+ * decayed to rest
+ */
+ list_move_tail(&flow->flowchain,
+ &b->decaying_flows);
+ if (flow->set == CAKE_SET_BULK) {
+ b->bulk_flow_count--;
+ b->decaying_flow_count++;
+ } else if (flow->set == CAKE_SET_SPARSE ||
+ flow->set == CAKE_SET_SPARSE_WAIT) {
+ b->sparse_flow_count--;
+ b->decaying_flow_count++;
+ }
+ flow->set = CAKE_SET_DECAYING;
+ } else {
+ /* remove empty queue from the flowchain */
+ list_del_init(&flow->flowchain);
+ if (flow->set == CAKE_SET_SPARSE ||
+ flow->set == CAKE_SET_SPARSE_WAIT)
+ b->sparse_flow_count--;
+ else if (flow->set == CAKE_SET_BULK)
+ b->bulk_flow_count--;
+ else
+ b->decaying_flow_count--;
+
+ flow->set = CAKE_SET_NONE;
+ srchost->srchost_refcnt--;
+ dsthost->dsthost_refcnt--;
+ }
+ goto begin;
+ }
+
+ /* Last packet in queue may be marked, shouldn't be dropped */
+ if (!cobalt_should_drop(&flow->cvars, &b->cparams, now, skb) ||
+ !flow->head)
+ break;
+
+ b->tin_dropped++;
+ qdisc_tree_reduce_backlog(sch, 1, qdisc_pkt_len(skb));
+ qdisc_qstats_drop(sch);
+ kfree_skb(skb);
+ }
+
+ b->tin_ecn_mark += !!flow->cvars.ecn_marked;
+ qdisc_bstats_update(sch, skb);
+
+ /* collect delay stats */
+ delay = now - cobalt_get_enqueue_time(skb);
+ b->avge_delay = cake_ewma(b->avge_delay, delay, 8);
+ b->peak_delay = cake_ewma(b->peak_delay, delay,
+ delay > b->peak_delay ? 2 : 8);
+ b->base_delay = cake_ewma(b->base_delay, delay,
+ delay < b->base_delay ? 2 : 8);
+
+ len = cake_advance_shaper(q, b, skb, now, false);
+ flow->deficit -= len;
+ b->tin_deficit -= len;
+
+ if (q->time_next_packet > now && sch->q.qlen) {
+ u64 next = min(q->time_next_packet, q->failsafe_next_packet);
+
+ qdisc_watchdog_schedule_ns(&q->watchdog, next);
+ } else if (!sch->q.qlen) {
+ int i;
+
+ for (i = 0; i < q->tin_cnt; i++) {
+ if (q->tins[i].decaying_flow_count) {
+ u64 next = now + q->tins[i].cparams.target;
+
+ qdisc_watchdog_schedule_ns(&q->watchdog, next);
+ break;
+ }
+ }
+ }
+
+ if (q->overflow_timeout)
+ q->overflow_timeout--;
+
+ return skb;
+}
+
+static void cake_reset(struct Qdisc *sch)
+{
+ u32 c;
+
+ for (c = 0; c < CAKE_MAX_TINS; c++)
+ cake_clear_tin(sch, c);
+}
+
+static const struct nla_policy cake_policy[TCA_CAKE_MAX + 1] = {
+ [TCA_CAKE_BASE_RATE64] = { .type = NLA_U64 },
+ [TCA_CAKE_DIFFSERV_MODE] = { .type = NLA_U32 },
+ [TCA_CAKE_ATM] = { .type = NLA_U32 },
+ [TCA_CAKE_FLOW_MODE] = { .type = NLA_U32 },
+ [TCA_CAKE_OVERHEAD] = { .type = NLA_S32 },
+ [TCA_CAKE_RTT] = { .type = NLA_U32 },
+ [TCA_CAKE_TARGET] = { .type = NLA_U32 },
+ [TCA_CAKE_AUTORATE] = { .type = NLA_U32 },
+ [TCA_CAKE_MEMORY] = { .type = NLA_U32 },
+ [TCA_CAKE_NAT] = { .type = NLA_U32 },
+ [TCA_CAKE_RAW] = { .type = NLA_U32 },
+ [TCA_CAKE_WASH] = { .type = NLA_U32 },
+ [TCA_CAKE_MPU] = { .type = NLA_U32 },
+ [TCA_CAKE_INGRESS] = { .type = NLA_U32 },
+ [TCA_CAKE_ACK_FILTER] = { .type = NLA_U32 },
+};
+
+static void cake_set_rate(struct cake_tin_data *b, u64 rate, u32 mtu,
+ cobalt_time_t ns_target, cobalt_time_t rtt_est_ns)
+{
+ /* convert byte-rate into time-per-byte
+ * so it will always unwedge in reasonable time.
+ */
+ static const u64 MIN_RATE = 64;
+ cobalt_time_t byte_target_ns;
+ u32 byte_target = mtu;
+ u8 rate_shft = 0;
+ u64 rate_ns = 0;
+
+ b->flow_quantum = 1514;
+ if (rate) {
+ b->flow_quantum = max(min(rate >> 12, 1514ULL), 300ULL);
+ rate_shft = 34;
+ rate_ns = ((u64)NSEC_PER_SEC) << rate_shft;
+ rate_ns /= max(MIN_RATE, rate);
+ while (!!(rate_ns >> 34)) {
+ rate_ns >>= 1;
+ rate_shft--;
+ }
+ } /* else unlimited, ie. zero delay */
+
+ b->tin_rate_bps = rate;
+ b->tin_rate_ns = rate_ns;
+ b->tin_rate_shft = rate_shft;
+
+ byte_target_ns = (byte_target * rate_ns) >> rate_shft;
+
+ b->cparams.target = max((byte_target_ns * 3) / 2, ns_target);
+ b->cparams.interval = max(rtt_est_ns +
+ b->cparams.target - ns_target,
+ b->cparams.target * 2);
+ b->cparams.mtu_time = byte_target_ns;
+ b->cparams.p_inc = 1 << 24; /* 1/256 */
+ b->cparams.p_dec = 1 << 20; /* 1/4096 */
+}
+
+static void cake_reconfigure(struct Qdisc *sch)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct cake_tin_data *b = &q->tins[0];
+ int c, ft = 0;
+
+ q->tin_cnt = 1;
+ cake_set_rate(b, q->rate_bps, psched_mtu(qdisc_dev(sch)),
+ US2TIME(q->target), US2TIME(q->interval));
+ b->tin_quantum_band = 65535;
+ b->tin_quantum_prio = 65535;
+
+ for (c = q->tin_cnt; c < CAKE_MAX_TINS; c++) {
+ cake_clear_tin(sch, c);
+ q->tins[c].cparams.mtu_time = q->tins[ft].cparams.mtu_time;
+ }
+
+ q->rate_ns = q->tins[ft].tin_rate_ns;
+ q->rate_shft = q->tins[ft].tin_rate_shft;
+
+ if (q->buffer_config_limit) {
+ q->buffer_limit = q->buffer_config_limit;
+ } else if (q->rate_bps) {
+ u64 t = q->rate_bps * q->interval;
+
+ do_div(t, USEC_PER_SEC / 4);
+ q->buffer_limit = max_t(u32, t, 4U << 20);
+ } else {
+ q->buffer_limit = ~0;
+ }
+
+ sch->flags &= ~TCQ_F_CAN_BYPASS;
+
+ q->buffer_limit = min(q->buffer_limit,
+ max(sch->limit * psched_mtu(qdisc_dev(sch)),
+ q->buffer_config_limit));
+}
+
+static int cake_change(struct Qdisc *sch, struct nlattr *opt,
+ struct netlink_ext_ack *extack)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct nlattr *tb[TCA_CAKE_MAX + 1];
+ int err;
+
+ if (!opt)
+ return -EINVAL;
+
+ err = nla_parse_nested(tb, TCA_CAKE_MAX, opt, cake_policy, extack);
+ if (err < 0)
+ return err;
+
+ if (tb[TCA_CAKE_BASE_RATE64])
+ q->rate_bps = nla_get_u64(tb[TCA_CAKE_BASE_RATE64]);
+
+ if (tb[TCA_CAKE_FLOW_MODE])
+ q->flow_mode = (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) &
+ CAKE_FLOW_MASK);
+
+ if (tb[TCA_CAKE_RTT]) {
+ q->interval = nla_get_u32(tb[TCA_CAKE_RTT]);
+
+ if (!q->interval)
+ q->interval = 1;
+ }
+
+ if (tb[TCA_CAKE_TARGET]) {
+ q->target = nla_get_u32(tb[TCA_CAKE_TARGET]);
+
+ if (!q->target)
+ q->target = 1;
+ }
+
+ if (tb[TCA_CAKE_MEMORY])
+ q->buffer_config_limit = nla_get_u32(tb[TCA_CAKE_MEMORY]);
+
+ if (q->tins) {
+ sch_tree_lock(sch);
+ cake_reconfigure(sch);
+ sch_tree_unlock(sch);
+ }
+
+ return 0;
+}
+
+static void cake_destroy(struct Qdisc *sch)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+
+ qdisc_watchdog_cancel(&q->watchdog);
+ tcf_block_put(q->block);
+ kvfree(q->tins);
+}
+
+static int cake_init(struct Qdisc *sch, struct nlattr *opt,
+ struct netlink_ext_ack *extack)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ int i, j, err;
+
+ sch->limit = 10240;
+ q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
+ q->flow_mode = CAKE_FLOW_TRIPLE;
+
+ q->rate_bps = 0; /* unlimited by default */
+
+ q->interval = 100000; /* 100ms default */
+ q->target = 5000; /* 5ms: codel RFC argues
+ * for 5 to 10% of interval
+ */
+
+ q->cur_tin = 0;
+ q->cur_flow = 0;
+
+ if (opt) {
+ int err = cake_change(sch, opt, extack);
+
+ if (err)
+ return err;
+ }
+
+ err = tcf_block_get(&q->block, &q->filter_list, sch, extack);
+ if (err)
+ return err;
+
+ qdisc_watchdog_init(&q->watchdog, sch);
+
+ quantum_div[0] = ~0;
+ for (i = 1; i <= CAKE_QUEUES; i++)
+ quantum_div[i] = 65535 / i;
+
+ q->tins = kvzalloc(CAKE_MAX_TINS * sizeof(struct cake_tin_data),
+ GFP_KERNEL);
+ if (!q->tins)
+ goto nomem;
+
+ for (i = 0; i < CAKE_MAX_TINS; i++) {
+ struct cake_tin_data *b = q->tins + i;
+
+ INIT_LIST_HEAD(&b->new_flows);
+ INIT_LIST_HEAD(&b->old_flows);
+ INIT_LIST_HEAD(&b->decaying_flows);
+ b->sparse_flow_count = 0;
+ b->bulk_flow_count = 0;
+ b->decaying_flow_count = 0;
+
+ for (j = 0; j < CAKE_QUEUES; j++) {
+ struct cake_flow *flow = b->flows + j;
+ u32 k = j * CAKE_MAX_TINS + i;
+
+ INIT_LIST_HEAD(&flow->flowchain);
+ cobalt_vars_init(&flow->cvars);
+
+ q->overflow_heap[k].t = i;
+ q->overflow_heap[k].b = j;
+ b->overflow_idx[j] = k;
+ }
+ }
+
+ cake_reconfigure(sch);
+ q->avg_peak_bandwidth = q->rate_bps;
+ q->min_netlen = ~0;
+ q->min_adjlen = ~0;
+ return 0;
+
+nomem:
+ cake_destroy(sch);
+ return -ENOMEM;
+}
+
+static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct nlattr *opts;
+
+ opts = nla_nest_start(skb, TCA_OPTIONS);
+ if (!opts)
+ goto nla_put_failure;
+
+ if (nla_put_u64_64bit(skb, TCA_CAKE_BASE_RATE64, q->rate_bps,
+ TCA_CAKE_TIN_STATS_PAD))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_FLOW_MODE,
+ q->flow_mode & CAKE_FLOW_MASK))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_RTT, q->interval))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_TARGET, q->target))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_MEMORY, q->buffer_config_limit))
+ goto nla_put_failure;
+
+ return nla_nest_end(skb, opts);
+
+nla_put_failure:
+ return -1;
+}
+
+static int cake_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
+{
+ struct nlattr *stats = nla_nest_start(d->skb, TCA_STATS_APP);
+ struct cake_sched_data *q = qdisc_priv(sch);
+ struct nlattr *tstats, *ts;
+ int i;
+
+ if (!stats)
+ return -1;
+
+#define PUT_STAT_U32(attr, data) do { \
+ if (nla_put_u32(d->skb, TCA_CAKE_STATS_ ## attr, data)) \
+ goto nla_put_failure; \
+ } while (0)
+#define PUT_STAT_U64(attr, data) do { \
+ if (nla_put_u64_64bit(d->skb, TCA_CAKE_STATS_ ## attr, \
+ data, TCA_CAKE_TIN_STATS_PAD)) \
+ goto nla_put_failure; \
+ } while (0)
+
+ PUT_STAT_U64(CAPACITY_ESTIMATE64, q->avg_peak_bandwidth);
+ PUT_STAT_U32(MEMORY_LIMIT, q->buffer_limit);
+ PUT_STAT_U32(MEMORY_USED, q->buffer_max_used);
+ PUT_STAT_U32(AVG_NETOFF, ((q->avg_netoff + 0x8000) >> 16));
+ PUT_STAT_U32(MAX_NETLEN, q->max_netlen);
+ PUT_STAT_U32(MAX_ADJLEN, q->max_adjlen);
+ PUT_STAT_U32(MIN_NETLEN, q->min_netlen);
+ PUT_STAT_U32(MIN_ADJLEN, q->min_adjlen);
+
+#undef PUT_STAT_U32
+#undef PUT_STAT_U64
+
+ tstats = nla_nest_start(d->skb, TCA_CAKE_STATS_TIN_STATS);
+ if (!tstats)
+ goto nla_put_failure;
+
+#define PUT_TSTAT_U32(attr, data) do { \
+ if (nla_put_u32(d->skb, TCA_CAKE_TIN_STATS_ ## attr, data)) \
+ goto nla_put_failure; \
+ } while (0)
+#define PUT_TSTAT_U64(attr, data) do { \
+ if (nla_put_u64_64bit(d->skb, TCA_CAKE_TIN_STATS_ ## attr, \
+ data, TCA_CAKE_TIN_STATS_PAD)) \
+ goto nla_put_failure; \
+ } while (0)
+
+ for (i = 0; i < q->tin_cnt; i++) {
+ struct cake_tin_data *b = &q->tins[i];
+
+ ts = nla_nest_start(d->skb, i + 1);
+ if (!ts)
+ goto nla_put_failure;
+
+ PUT_TSTAT_U64(THRESHOLD_RATE64, b->tin_rate_bps);
+ PUT_TSTAT_U32(TARGET_US, cobalt_time_to_us(b->cparams.target));
+ PUT_TSTAT_U32(INTERVAL_US,
+ cobalt_time_to_us(b->cparams.interval));
+
+ PUT_TSTAT_U32(SENT_PACKETS, b->packets);
+ PUT_TSTAT_U64(SENT_BYTES64, b->bytes);
+ PUT_TSTAT_U32(DROPPED_PACKETS, b->tin_dropped);
+ PUT_TSTAT_U32(ECN_MARKED_PACKETS, b->tin_ecn_mark);
+ PUT_TSTAT_U64(BACKLOG_BYTES64, b->tin_backlog);
+ PUT_TSTAT_U32(ACKS_DROPPED_PACKETS, b->ack_drops);
+
+ PUT_TSTAT_U32(PEAK_DELAY_US, cobalt_time_to_us(b->peak_delay));
+ PUT_TSTAT_U32(AVG_DELAY_US, cobalt_time_to_us(b->avge_delay));
+ PUT_TSTAT_U32(BASE_DELAY_US, cobalt_time_to_us(b->base_delay));
+
+ PUT_TSTAT_U32(WAY_INDIRECT_HITS, b->way_hits);
+ PUT_TSTAT_U32(WAY_MISSES, b->way_misses);
+ PUT_TSTAT_U32(WAY_COLLISIONS, b->way_collisions);
+
+ PUT_TSTAT_U32(SPARSE_FLOWS, b->sparse_flow_count +
+ b->decaying_flow_count);
+ PUT_TSTAT_U32(BULK_FLOWS, b->bulk_flow_count);
+ PUT_TSTAT_U32(UNRESPONSIVE_FLOWS, b->unresponsive_flow_count);
+ PUT_TSTAT_U32(MAX_SKBLEN, b->max_skblen);
+
+ PUT_TSTAT_U32(FLOW_QUANTUM, b->flow_quantum);
+ nla_nest_end(d->skb, ts);
+ }
+
+#undef PUT_TSTAT_U32
+#undef PUT_TSTAT_U64
+
+ nla_nest_end(d->skb, tstats);
+ return nla_nest_end(d->skb, stats);
+
+nla_put_failure:
+ nla_nest_cancel(d->skb, stats);
+ return -1;
+}
+
+static struct Qdisc *cake_leaf(struct Qdisc *sch, unsigned long arg)
+{
+ return NULL;
+}
+
+static unsigned long cake_find(struct Qdisc *sch, u32 classid)
+{
+ return 0;
+}
+
+static void cake_walk(struct Qdisc *sch, struct qdisc_walker *arg)
+{
+}
+
+static unsigned long cake_bind(struct Qdisc *sch, unsigned long parent,
+ u32 classid)
+{
+ return 0;
+}
+
+static void cake_unbind(struct Qdisc *q, unsigned long cl)
+{
+}
+
+static struct tcf_block *cake_tcf_block(struct Qdisc *sch, unsigned long cl,
+ struct netlink_ext_ack *extack)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+
+ if (cl)
+ return NULL;
+ return q->block;
+}
+
+static const struct Qdisc_class_ops cake_class_ops = {
+ .leaf = cake_leaf,
+ .find = cake_find,
+ .tcf_block = cake_tcf_block,
+ .bind_tcf = cake_bind,
+ .unbind_tcf = cake_unbind,
+ .walk = cake_walk,
+};
+
+static struct Qdisc_ops cake_qdisc_ops __read_mostly = {
+ .cl_ops = &cake_class_ops,
+ .id = "cake",
+ .priv_size = sizeof(struct cake_sched_data),
+ .enqueue = cake_enqueue,
+ .dequeue = cake_dequeue,
+ .peek = qdisc_peek_dequeued,
+ .init = cake_init,
+ .reset = cake_reset,
+ .destroy = cake_destroy,
+ .change = cake_change,
+ .dump = cake_dump,
+ .dump_stats = cake_dump_stats,
+ .owner = THIS_MODULE,
+};
+
+static int __init cake_module_init(void)
+{
+ return register_qdisc(&cake_qdisc_ops);
+}
+
+static void __exit cake_module_exit(void)
+{
+ unregister_qdisc(&cake_qdisc_ops);
+}
+
+module_init(cake_module_init)
+module_exit(cake_module_exit)
+MODULE_AUTHOR("Jonathan Morton");
+MODULE_LICENSE("Dual BSD/GPL");
+MODULE_DESCRIPTION("The CAKE shaper.");
^ permalink raw reply related
* [PATCH net-next v11 5/7] sch_cake: Add DiffServ handling
From: Toke Høiland-Jørgensen @ 2018-05-15 15:12 UTC (permalink / raw)
To: netdev; +Cc: cake
In-Reply-To: <152639705919.2525.3718182212727523960.stgit@alrua-kau>
This adds support for DiffServ-based priority queueing to CAKE. If the
shaper is in use, each priority tier gets its own virtual clock, which
limits that tier's rate to a fraction of the overall shaped rate, to
discourage trying to game the priority mechanism.
CAKE defaults to a simple, three-tier mode that interprets most code points
as "best effort", but places CS1 traffic into a low-priority "bulk" tier
which is assigned 1/16 of the total rate, and a few code points indicating
latency-sensitive or control traffic (specifically TOS4, VA, EF, CS6, CS7)
into a "latency sensitive" high-priority tier, which is assigned 1/4 rate.
The other supported DiffServ modes are a 4-tier mode matching the 802.11e
precedence rules, as well as two 8-tier modes, one of which implements
strict precedence of the eight priority levels.
This commit also adds an optional DiffServ 'wash' mode, which will zero out
the DSCP fields of any packet passing through CAKE. While this can
technically be done with other mechanisms in the kernel, having the feature
available in CAKE significantly decreases configuration complexity; and the
implementation cost is low on top of the other DiffServ-handling code.
Filters and applications can set the skb->priority field to override the
DSCP-based classification into tiers. If TC_H_MAJ(skb->priority) matches
CAKE's qdisc handle, the minor number will be interpreted as a priority
tier if it is less than or equal to the number of configured priority
tiers.
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
---
net/sched/sch_cake.c | 408 +++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 401 insertions(+), 7 deletions(-)
diff --git a/net/sched/sch_cake.c b/net/sched/sch_cake.c
index 14622c1d970e..4b13345f85dc 100644
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -307,6 +307,68 @@ static void cobalt_set_enqueue_time(struct sk_buff *skb,
static u16 quantum_div[CAKE_QUEUES + 1] = {0};
+/* Diffserv lookup tables */
+
+static const u8 precedence[] = {
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 1, 1, 1, 1, 1, 1, 1, 1,
+ 2, 2, 2, 2, 2, 2, 2, 2,
+ 3, 3, 3, 3, 3, 3, 3, 3,
+ 4, 4, 4, 4, 4, 4, 4, 4,
+ 5, 5, 5, 5, 5, 5, 5, 5,
+ 6, 6, 6, 6, 6, 6, 6, 6,
+ 7, 7, 7, 7, 7, 7, 7, 7,
+};
+
+static const u8 diffserv8[] = {
+ 2, 5, 1, 2, 4, 2, 2, 2,
+ 0, 2, 1, 2, 1, 2, 1, 2,
+ 5, 2, 4, 2, 4, 2, 4, 2,
+ 3, 2, 3, 2, 3, 2, 3, 2,
+ 6, 2, 3, 2, 3, 2, 3, 2,
+ 6, 2, 2, 2, 6, 2, 6, 2,
+ 7, 2, 2, 2, 2, 2, 2, 2,
+ 7, 2, 2, 2, 2, 2, 2, 2,
+};
+
+static const u8 diffserv4[] = {
+ 0, 2, 0, 0, 2, 0, 0, 0,
+ 1, 0, 0, 0, 0, 0, 0, 0,
+ 2, 0, 2, 0, 2, 0, 2, 0,
+ 2, 0, 2, 0, 2, 0, 2, 0,
+ 3, 0, 2, 0, 2, 0, 2, 0,
+ 3, 0, 0, 0, 3, 0, 3, 0,
+ 3, 0, 0, 0, 0, 0, 0, 0,
+ 3, 0, 0, 0, 0, 0, 0, 0,
+};
+
+static const u8 diffserv3[] = {
+ 0, 0, 0, 0, 2, 0, 0, 0,
+ 1, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 2, 0, 2, 0,
+ 2, 0, 0, 0, 0, 0, 0, 0,
+ 2, 0, 0, 0, 0, 0, 0, 0,
+};
+
+static const u8 besteffort[] = {
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0,
+};
+
+/* tin priority order for stats dumping */
+
+static const u8 normal_order[] = {0, 1, 2, 3, 4, 5, 6, 7};
+static const u8 bulk_order[] = {1, 0, 2, 3};
+
#define REC_INV_SQRT_CACHE (16)
static u32 cobalt_rec_inv_sqrt_cache[REC_INV_SQRT_CACHE] = {0};
@@ -1223,6 +1285,46 @@ static unsigned int cake_drop(struct Qdisc *sch, struct sk_buff **to_free)
return idx + (tin << 16);
}
+static void cake_wash_diffserv(struct sk_buff *skb)
+{
+ switch (skb->protocol) {
+ case htons(ETH_P_IP):
+ ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, 0);
+ break;
+ case htons(ETH_P_IPV6):
+ ipv6_change_dsfield(ipv6_hdr(skb), INET_ECN_MASK, 0);
+ break;
+ default:
+ break;
+ }
+}
+
+static u8 cake_handle_diffserv(struct sk_buff *skb, u16 wash)
+{
+ u8 dscp;
+
+ switch (skb->protocol) {
+ case htons(ETH_P_IP):
+ dscp = ipv4_get_dsfield(ip_hdr(skb)) >> 2;
+ if (wash && dscp)
+ ipv4_change_dsfield(ip_hdr(skb), INET_ECN_MASK, 0);
+ return dscp;
+
+ case htons(ETH_P_IPV6):
+ dscp = ipv6_get_dsfield(ipv6_hdr(skb)) >> 2;
+ if (wash && dscp)
+ ipv6_change_dsfield(ipv6_hdr(skb), INET_ECN_MASK, 0);
+ return dscp;
+
+ case htons(ETH_P_ARP):
+ return 0x38; /* CS7 - Net Control */
+
+ default:
+ /* If there is no Diffserv field, treat as best-effort */
+ return 0;
+ }
+}
+
static void cake_reconfigure(struct Qdisc *sch);
static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
@@ -1237,7 +1339,26 @@ static s32 cake_enqueue(struct sk_buff *skb, struct Qdisc *sch,
struct cake_flow *flow;
u32 idx, tin;
- tin = 0;
+ if (TC_H_MAJ(skb->priority) == sch->handle &&
+ TC_H_MIN(skb->priority) > 0 &&
+ TC_H_MIN(skb->priority) <= q->tin_cnt) {
+ tin = TC_H_MIN(skb->priority) - 1;
+
+ if (q->rate_flags & CAKE_FLAG_WASH)
+ cake_wash_diffserv(skb);
+ } else if (q->tin_mode != CAKE_DIFFSERV_BESTEFFORT) {
+ /* extract the Diffserv Precedence field, if it exists */
+ /* and clear DSCP bits if washing */
+ tin = q->tin_index[cake_handle_diffserv(skb,
+ q->rate_flags & CAKE_FLAG_WASH)];
+ if (unlikely(tin >= q->tin_cnt))
+ tin = 0;
+ } else {
+ tin = 0;
+ if (q->rate_flags & CAKE_FLAG_WASH)
+ cake_wash_diffserv(skb);
+ }
+
b = &q->tins[tin];
/* choose flow to insert into */
@@ -1719,18 +1840,274 @@ static void cake_set_rate(struct cake_tin_data *b, u64 rate, u32 mtu,
b->cparams.p_dec = 1 << 20; /* 1/4096 */
}
-static void cake_reconfigure(struct Qdisc *sch)
+static int cake_config_besteffort(struct Qdisc *sch)
{
struct cake_sched_data *q = qdisc_priv(sch);
struct cake_tin_data *b = &q->tins[0];
- int c, ft = 0;
+ u32 mtu = psched_mtu(qdisc_dev(sch));
+ u64 rate = q->rate_bps;
q->tin_cnt = 1;
- cake_set_rate(b, q->rate_bps, psched_mtu(qdisc_dev(sch)),
- US2TIME(q->target), US2TIME(q->interval));
+
+ q->tin_index = besteffort;
+ q->tin_order = normal_order;
+
+ cake_set_rate(b, rate, mtu, US2TIME(q->target), US2TIME(q->interval));
b->tin_quantum_band = 65535;
b->tin_quantum_prio = 65535;
+ return 0;
+}
+
+static int cake_config_precedence(struct Qdisc *sch)
+{
+ /* convert high-level (user visible) parameters into internal format */
+ struct cake_sched_data *q = qdisc_priv(sch);
+ u32 mtu = psched_mtu(qdisc_dev(sch));
+ u64 rate = q->rate_bps;
+ u32 quantum1 = 256;
+ u32 quantum2 = 256;
+ u32 i;
+
+ q->tin_cnt = 8;
+ q->tin_index = precedence;
+ q->tin_order = normal_order;
+
+ for (i = 0; i < q->tin_cnt; i++) {
+ struct cake_tin_data *b = &q->tins[i];
+
+ cake_set_rate(b, rate, mtu, US2TIME(q->target),
+ US2TIME(q->interval));
+
+ b->tin_quantum_prio = max_t(u16, 1U, quantum1);
+ b->tin_quantum_band = max_t(u16, 1U, quantum2);
+
+ /* calculate next class's parameters */
+ rate *= 7;
+ rate >>= 3;
+
+ quantum1 *= 3;
+ quantum1 >>= 1;
+
+ quantum2 *= 7;
+ quantum2 >>= 3;
+ }
+
+ return 0;
+}
+
+/* List of known Diffserv codepoints:
+ *
+ * Least Effort (CS1)
+ * Best Effort (CS0)
+ * Max Reliability & LLT "Lo" (TOS1)
+ * Max Throughput (TOS2)
+ * Min Delay (TOS4)
+ * LLT "La" (TOS5)
+ * Assured Forwarding 1 (AF1x) - x3
+ * Assured Forwarding 2 (AF2x) - x3
+ * Assured Forwarding 3 (AF3x) - x3
+ * Assured Forwarding 4 (AF4x) - x3
+ * Precedence Class 2 (CS2)
+ * Precedence Class 3 (CS3)
+ * Precedence Class 4 (CS4)
+ * Precedence Class 5 (CS5)
+ * Precedence Class 6 (CS6)
+ * Precedence Class 7 (CS7)
+ * Voice Admit (VA)
+ * Expedited Forwarding (EF)
+
+ * Total 25 codepoints.
+ */
+
+/* List of traffic classes in RFC 4594:
+ * (roughly descending order of contended priority)
+ * (roughly ascending order of uncontended throughput)
+ *
+ * Network Control (CS6,CS7) - routing traffic
+ * Telephony (EF,VA) - aka. VoIP streams
+ * Signalling (CS5) - VoIP setup
+ * Multimedia Conferencing (AF4x) - aka. video calls
+ * Realtime Interactive (CS4) - eg. games
+ * Multimedia Streaming (AF3x) - eg. YouTube, NetFlix, Twitch
+ * Broadcast Video (CS3)
+ * Low Latency Data (AF2x,TOS4) - eg. database
+ * Ops, Admin, Management (CS2,TOS1) - eg. ssh
+ * Standard Service (CS0 & unrecognised codepoints)
+ * High Throughput Data (AF1x,TOS2) - eg. web traffic
+ * Low Priority Data (CS1) - eg. BitTorrent
+
+ * Total 12 traffic classes.
+ */
+
+static int cake_config_diffserv8(struct Qdisc *sch)
+{
+/* Pruned list of traffic classes for typical applications:
+ *
+ * Network Control (CS6, CS7)
+ * Minimum Latency (EF, VA, CS5, CS4)
+ * Interactive Shell (CS2, TOS1)
+ * Low Latency Transactions (AF2x, TOS4)
+ * Video Streaming (AF4x, AF3x, CS3)
+ * Bog Standard (CS0 etc.)
+ * High Throughput (AF1x, TOS2)
+ * Background Traffic (CS1)
+ *
+ * Total 8 traffic classes.
+ */
+
+ struct cake_sched_data *q = qdisc_priv(sch);
+ u32 mtu = psched_mtu(qdisc_dev(sch));
+ u64 rate = q->rate_bps;
+ u32 quantum1 = 256;
+ u32 quantum2 = 256;
+ u32 i;
+
+ q->tin_cnt = 8;
+
+ /* codepoint to class mapping */
+ q->tin_index = diffserv8;
+ q->tin_order = normal_order;
+
+ /* class characteristics */
+ for (i = 0; i < q->tin_cnt; i++) {
+ struct cake_tin_data *b = &q->tins[i];
+
+ cake_set_rate(b, rate, mtu, US2TIME(q->target),
+ US2TIME(q->interval));
+
+ b->tin_quantum_prio = max_t(u16, 1U, quantum1);
+ b->tin_quantum_band = max_t(u16, 1U, quantum2);
+
+ /* calculate next class's parameters */
+ rate *= 7;
+ rate >>= 3;
+
+ quantum1 *= 3;
+ quantum1 >>= 1;
+
+ quantum2 *= 7;
+ quantum2 >>= 3;
+ }
+
+ return 0;
+}
+
+static int cake_config_diffserv4(struct Qdisc *sch)
+{
+/* Further pruned list of traffic classes for four-class system:
+ *
+ * Latency Sensitive (CS7, CS6, EF, VA, CS5, CS4)
+ * Streaming Media (AF4x, AF3x, CS3, AF2x, TOS4, CS2, TOS1)
+ * Best Effort (CS0, AF1x, TOS2, and those not specified)
+ * Background Traffic (CS1)
+ *
+ * Total 4 traffic classes.
+ */
+
+ struct cake_sched_data *q = qdisc_priv(sch);
+ u32 mtu = psched_mtu(qdisc_dev(sch));
+ u64 rate = q->rate_bps;
+ u32 quantum = 1024;
+
+ q->tin_cnt = 4;
+
+ /* codepoint to class mapping */
+ q->tin_index = diffserv4;
+ q->tin_order = bulk_order;
+
+ /* class characteristics */
+ cake_set_rate(&q->tins[0], rate, mtu,
+ US2TIME(q->target), US2TIME(q->interval));
+ cake_set_rate(&q->tins[1], rate >> 4, mtu,
+ US2TIME(q->target), US2TIME(q->interval));
+ cake_set_rate(&q->tins[2], rate >> 1, mtu,
+ US2TIME(q->target), US2TIME(q->interval));
+ cake_set_rate(&q->tins[3], rate >> 2, mtu,
+ US2TIME(q->target), US2TIME(q->interval));
+
+ /* priority weights */
+ q->tins[0].tin_quantum_prio = quantum;
+ q->tins[1].tin_quantum_prio = quantum >> 4;
+ q->tins[2].tin_quantum_prio = quantum << 2;
+ q->tins[3].tin_quantum_prio = quantum << 4;
+
+ /* bandwidth-sharing weights */
+ q->tins[0].tin_quantum_band = quantum;
+ q->tins[1].tin_quantum_band = quantum >> 4;
+ q->tins[2].tin_quantum_band = quantum >> 1;
+ q->tins[3].tin_quantum_band = quantum >> 2;
+
+ return 0;
+}
+
+static int cake_config_diffserv3(struct Qdisc *sch)
+{
+/* Simplified Diffserv structure with 3 tins.
+ * Low Priority (CS1)
+ * Best Effort
+ * Latency Sensitive (TOS4, VA, EF, CS6, CS7)
+ */
+ struct cake_sched_data *q = qdisc_priv(sch);
+ u32 mtu = psched_mtu(qdisc_dev(sch));
+ u64 rate = q->rate_bps;
+ u32 quantum = 1024;
+
+ q->tin_cnt = 3;
+
+ /* codepoint to class mapping */
+ q->tin_index = diffserv3;
+ q->tin_order = bulk_order;
+
+ /* class characteristics */
+ cake_set_rate(&q->tins[0], rate, mtu,
+ US2TIME(q->target), US2TIME(q->interval));
+ cake_set_rate(&q->tins[1], rate >> 4, mtu,
+ US2TIME(q->target), US2TIME(q->interval));
+ cake_set_rate(&q->tins[2], rate >> 2, mtu,
+ US2TIME(q->target), US2TIME(q->interval));
+
+ /* priority weights */
+ q->tins[0].tin_quantum_prio = quantum;
+ q->tins[1].tin_quantum_prio = quantum >> 4;
+ q->tins[2].tin_quantum_prio = quantum << 4;
+
+ /* bandwidth-sharing weights */
+ q->tins[0].tin_quantum_band = quantum;
+ q->tins[1].tin_quantum_band = quantum >> 4;
+ q->tins[2].tin_quantum_band = quantum >> 2;
+
+ return 0;
+}
+
+static void cake_reconfigure(struct Qdisc *sch)
+{
+ struct cake_sched_data *q = qdisc_priv(sch);
+ int c, ft;
+
+ switch (q->tin_mode) {
+ case CAKE_DIFFSERV_BESTEFFORT:
+ ft = cake_config_besteffort(sch);
+ break;
+
+ case CAKE_DIFFSERV_PRECEDENCE:
+ ft = cake_config_precedence(sch);
+ break;
+
+ case CAKE_DIFFSERV_DIFFSERV8:
+ ft = cake_config_diffserv8(sch);
+ break;
+
+ case CAKE_DIFFSERV_DIFFSERV4:
+ ft = cake_config_diffserv4(sch);
+ break;
+
+ case CAKE_DIFFSERV_DIFFSERV3:
+ default:
+ ft = cake_config_diffserv3(sch);
+ break;
+ }
+
for (c = q->tin_cnt; c < CAKE_MAX_TINS; c++) {
cake_clear_tin(sch, c);
q->tins[c].cparams.mtu_time = q->tins[ft].cparams.mtu_time;
@@ -1774,6 +2151,16 @@ static int cake_change(struct Qdisc *sch, struct nlattr *opt,
if (tb[TCA_CAKE_BASE_RATE64])
q->rate_bps = nla_get_u64(tb[TCA_CAKE_BASE_RATE64]);
+ if (tb[TCA_CAKE_DIFFSERV_MODE])
+ q->tin_mode = nla_get_u32(tb[TCA_CAKE_DIFFSERV_MODE]);
+
+ if (tb[TCA_CAKE_WASH]) {
+ if (!!nla_get_u32(tb[TCA_CAKE_WASH]))
+ q->rate_flags |= CAKE_FLAG_WASH;
+ else
+ q->rate_flags &= ~CAKE_FLAG_WASH;
+ }
+
if (tb[TCA_CAKE_FLOW_MODE])
q->flow_mode = (nla_get_u32(tb[TCA_CAKE_FLOW_MODE]) &
CAKE_FLOW_MASK);
@@ -1843,7 +2230,7 @@ static int cake_init(struct Qdisc *sch, struct nlattr *opt,
int i, j, err;
sch->limit = 10240;
- q->tin_mode = CAKE_DIFFSERV_BESTEFFORT;
+ q->tin_mode = CAKE_DIFFSERV_DIFFSERV3;
q->flow_mode = CAKE_FLOW_TRIPLE;
q->rate_bps = 0; /* unlimited by default */
@@ -1953,6 +2340,13 @@ static int cake_dump(struct Qdisc *sch, struct sk_buff *skb)
!!(q->flow_mode & CAKE_FLOW_NAT_FLAG)))
goto nla_put_failure;
+ if (nla_put_u32(skb, TCA_CAKE_DIFFSERV_MODE, q->tin_mode))
+ goto nla_put_failure;
+
+ if (nla_put_u32(skb, TCA_CAKE_WASH,
+ !!(q->rate_flags & CAKE_FLAG_WASH)))
+ goto nla_put_failure;
+
return nla_nest_end(skb, opts);
nla_put_failure:
@@ -2006,7 +2400,7 @@ static int cake_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
} while (0)
for (i = 0; i < q->tin_cnt; i++) {
- struct cake_tin_data *b = &q->tins[i];
+ struct cake_tin_data *b = &q->tins[q->tin_order[i]];
ts = nla_nest_start(d->skb, i + 1);
if (!ts)
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox