* Re: [patch 1/1] drivers/net/ethernet/marvell/mvmdio.c: Fix non OF case
From: Andrew Lunn @ 2019-08-03 22:22 UTC (permalink / raw)
To: Arnaud Patard; +Cc: netdev
In-Reply-To: <20190802083310.772136040@rtp-net.org>
On Fri, Aug 02, 2019 at 10:32:40AM +0200, Arnaud Patard wrote:
> Orion5.x systems are still using machine files and not device-tree.
> Commit 96cb4342382290c9 ("net: mvmdio: allow up to three clocks to be
> specified for orion-mdio") has replaced devm_clk_get() with of_clk_get(),
> leading to a oops at boot and not working network, as reported in
> https://lists.debian.org/debian-arm/2019/07/msg00088.html and possibly in
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=908712.
>
> Link: https://lists.debian.org/debian-arm/2019/07/msg00088.html
> Fixes: 96cb4342382290c9 ("net: mvmdio: allow up to three clocks to be specified for orion-mdio")
> Signed-off-by: Arnaud Patard <arnaud.patard@rtp-net.org>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Andrew
^ permalink raw reply
* Re: [PATCH net-next] openvswitch: Print error when ovs_execute_actions() fails
From: Pravin Shelar @ 2019-08-03 23:01 UTC (permalink / raw)
To: Yifeng Sun; +Cc: Linux Kernel Network Developers
In-Reply-To: <1564694047-4859-1-git-send-email-pkusunyifeng@gmail.com>
On Thu, Aug 1, 2019 at 2:14 PM Yifeng Sun <pkusunyifeng@gmail.com> wrote:
>
> Currently in function ovs_dp_process_packet(), return values of
> ovs_execute_actions() are silently discarded. This patch prints out
> an error message when error happens so as to provide helpful hints
> for debugging.
>
> Signed-off-by: Yifeng Sun <pkusunyifeng@gmail.com>
> ---
> net/openvswitch/datapath.c | 7 +++++--
> 1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/net/openvswitch/datapath.c b/net/openvswitch/datapath.c
> index 892287d..603c533 100644
> --- a/net/openvswitch/datapath.c
> +++ b/net/openvswitch/datapath.c
> @@ -222,6 +222,7 @@ void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
> struct dp_stats_percpu *stats;
> u64 *stats_counter;
> u32 n_mask_hit;
> + int error;
>
> stats = this_cpu_ptr(dp->stats_percpu);
>
> @@ -229,7 +230,6 @@ void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
> flow = ovs_flow_tbl_lookup_stats(&dp->table, key, &n_mask_hit);
> if (unlikely(!flow)) {
> struct dp_upcall_info upcall;
> - int error;
>
> memset(&upcall, 0, sizeof(upcall));
> upcall.cmd = OVS_PACKET_CMD_MISS;
> @@ -246,7 +246,10 @@ void ovs_dp_process_packet(struct sk_buff *skb, struct sw_flow_key *key)
>
> ovs_flow_stats_update(flow, key->tp.flags, skb);
> sf_acts = rcu_dereference(flow->sf_acts);
> - ovs_execute_actions(dp, skb, sf_acts, key);
> + error = ovs_execute_actions(dp, skb, sf_acts, key);
> + if (unlikely(error))
> + net_err_ratelimited("ovs: action execution error on datapath %s: %d\n",
> + ovs_dp_name(dp), error);
>
I would rather add error counter for better visibility.
If you want to use current approach, can you use net_dbg_ratelimited()
since you want to use this for debugging purpose?
Thanks.
^ permalink raw reply
* [PATCH net-next 0/2] Two small fq_codel optimizations
From: Dave Taht @ 2019-08-03 23:37 UTC (permalink / raw)
To: netdev; +Cc: Dave Taht
These two patches improve fq_codel performance
under extreme network loads. The first patch
more rapidly escalates the codel count under
overload, the second just kills a totally useless
statistic.
(sent together because they'd otherwise conflict)
Signed-off-by: Dave Taht <dave.taht@gmail.com>
Dave Taht (2):
Increase fq_codel count in the bulk dropper
fq_codel: Kill useless per-flow dropped statistic
net/sched/sch_fq_codel.c | 10 +++-------
1 file changed, 3 insertions(+), 7 deletions(-)
--
2.17.1
^ permalink raw reply
* [PATCH net-next 1/2] Increase fq_codel count in the bulk dropper
From: Dave Taht @ 2019-08-03 23:37 UTC (permalink / raw)
To: netdev; +Cc: Dave Taht
In-Reply-To: <1564875449-12122-1-git-send-email-dave.taht@gmail.com>
In the field fq_codel is often used with a smaller memory or
packet limit than the default, and when the bulk dropper is hit,
the drop pattern bifircates into one that more slowly increases
the codel drop rate and hits the bulk dropper more than it should.
The scan through the 1024 queues happens more often than it needs to.
This patch increases the codel count in the bulk dropper, but
does not change the drop rate there, relying on the next codel round
to deliver the next packet at the original drop rate
(after that burst of loss), then escalate to a higher signaling rate.
Signed-off-by: Dave Taht <dave.taht@gmail.com>
---
net/sched/sch_fq_codel.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/net/sched/sch_fq_codel.c b/net/sched/sch_fq_codel.c
index d59fbcc745d1..d67b2c40e6e6 100644
--- a/net/sched/sch_fq_codel.c
+++ b/net/sched/sch_fq_codel.c
@@ -173,6 +173,8 @@ static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets,
__qdisc_drop(skb, to_free);
} while (++i < max_packets && len < threshold);
+ /* Tell codel to increase its signal strength also */
+ flow->cvars.count += i;
flow->dropped += i;
q->backlogs[idx] -= len;
q->memory_usage -= mem;
--
2.17.1
^ permalink raw reply related
* [PATCH net-next 2/2] fq_codel: Kill useless per-flow dropped statistic
From: Dave Taht @ 2019-08-03 23:37 UTC (permalink / raw)
To: netdev; +Cc: Dave Taht
In-Reply-To: <1564875449-12122-1-git-send-email-dave.taht@gmail.com>
It is almost impossible to get anything other than a 0 out of
flow->dropped statistic with a tc class dump, as it resets to 0
on every round.
It also conflates ecn marks with drops.
It would have been useful had it kept a cumulative drop count, but
it doesn't. This patch doesn't change the API, it just stops
tracking a stat and state that is impossible to measure and nobody
uses.
Signed-off-by: Dave Taht <dave.taht@gmail.com>
---
net/sched/sch_fq_codel.c | 8 +-------
1 file changed, 1 insertion(+), 7 deletions(-)
diff --git a/net/sched/sch_fq_codel.c b/net/sched/sch_fq_codel.c
index d67b2c40e6e6..9edd0f495001 100644
--- a/net/sched/sch_fq_codel.c
+++ b/net/sched/sch_fq_codel.c
@@ -45,7 +45,6 @@ struct fq_codel_flow {
struct sk_buff *tail;
struct list_head flowchain;
int deficit;
- u32 dropped; /* number of drops (or ECN marks) on this flow */
struct codel_vars cvars;
}; /* please try to keep this structure <= 64 bytes */
@@ -175,7 +174,6 @@ static unsigned int fq_codel_drop(struct Qdisc *sch, unsigned int max_packets,
/* Tell codel to increase its signal strength also */
flow->cvars.count += i;
- flow->dropped += i;
q->backlogs[idx] -= len;
q->memory_usage -= mem;
sch->qstats.drops += i;
@@ -213,7 +211,6 @@ static int fq_codel_enqueue(struct sk_buff *skb, struct Qdisc *sch,
list_add_tail(&flow->flowchain, &q->new_flows);
q->new_flow_count++;
flow->deficit = q->quantum;
- flow->dropped = 0;
}
get_codel_cb(skb)->mem_usage = skb->truesize;
q->memory_usage += get_codel_cb(skb)->mem_usage;
@@ -312,9 +309,6 @@ static struct sk_buff *fq_codel_dequeue(struct Qdisc *sch)
&flow->cvars, &q->cstats, qdisc_pkt_len,
codel_get_enqueue_time, drop_func, dequeue_func);
- flow->dropped += q->cstats.drop_count - prev_drop_count;
- flow->dropped += q->cstats.ecn_mark - prev_ecn_mark;
-
if (!skb) {
/* force a pass through old_flows to prevent starvation */
if ((head == &q->new_flows) && !list_empty(&q->old_flows))
@@ -660,7 +654,7 @@ static int fq_codel_dump_class_stats(struct Qdisc *sch, unsigned long cl,
sch_tree_unlock(sch);
}
qs.backlog = q->backlogs[idx];
- qs.drops = flow->dropped;
+ qs.drops = 0;
}
if (gnet_stats_copy_queue(d, NULL, &qs, qs.qlen) < 0)
return -1;
--
2.17.1
^ permalink raw reply related
* Re: [PATCH V2 7/9] vhost: do not use RCU to synchronize MMU notifier with worker
From: Jason Gunthorpe @ 2019-08-04 0:14 UTC (permalink / raw)
To: Michael S. Tsirkin
Cc: Jason Wang, kvm, virtualization, netdev, linux-kernel, linux-mm
In-Reply-To: <20190803172944-mutt-send-email-mst@kernel.org>
On Sat, Aug 03, 2019 at 05:36:13PM -0400, Michael S. Tsirkin wrote:
> On Fri, Aug 02, 2019 at 02:24:18PM -0300, Jason Gunthorpe wrote:
> > On Fri, Aug 02, 2019 at 10:27:21AM -0400, Michael S. Tsirkin wrote:
> > > On Fri, Aug 02, 2019 at 09:46:13AM -0300, Jason Gunthorpe wrote:
> > > > On Fri, Aug 02, 2019 at 05:40:07PM +0800, Jason Wang wrote:
> > > > > > This must be a proper barrier, like a spinlock, mutex, or
> > > > > > synchronize_rcu.
> > > > >
> > > > >
> > > > > I start with synchronize_rcu() but both you and Michael raise some
> > > > > concern.
> > > >
> > > > I've also idly wondered if calling synchronize_rcu() under the various
> > > > mm locks is a deadlock situation.
> > > >
> > > > > Then I try spinlock and mutex:
> > > > >
> > > > > 1) spinlock: add lots of overhead on datapath, this leads 0 performance
> > > > > improvement.
> > > >
> > > > I think the topic here is correctness not performance improvement
> > >
> > > The topic is whether we should revert
> > > commit 7f466032dc9 ("vhost: access vq metadata through kernel virtual address")
> > >
> > > or keep it in. The only reason to keep it is performance.
> >
> > Yikes, I'm not sure you can ever win against copy_from_user using
> > mmu_notifiers?
>
> Ever since copy_from_user started playing with flags (for SMAP) and
> added speculation barriers there's a chance we can win by accessing
> memory through the kernel address.
You think copy_to_user will be more expensive than the minimum two
atomics required to synchronize with another thread?
> > Also, why can't this just permanently GUP the pages? In fact, where
> > does it put_page them anyhow? Worrying that 7f466 adds a get_user page
> > but does not add a put_page??
You didn't answer this.. Why not just use GUP?
Jason
^ permalink raw reply
* [PATCH 1/2] Fix a NULL-ptr-deref bug in ath6kl_usb_alloc_urb_from_pipe
From: Hui Peng @ 2019-08-04 0:29 UTC (permalink / raw)
To: kvalo, davem
Cc: Hui Peng, Mathias Payer, linux-wireless, netdev, linux-kernel
The `ar_usb` field of `ath6kl_usb_pipe_usb_pipe` objects
are initialized to point to the containing `ath6kl_usb` object
according to endpoint descriptors read from the device side, as shown
below in `ath6kl_usb_setup_pipe_resources`:
for (i = 0; i < iface_desc->desc.bNumEndpoints; ++i) {
endpoint = &iface_desc->endpoint[i].desc;
// get the address from endpoint descriptor
pipe_num = ath6kl_usb_get_logical_pipe_num(ar_usb,
endpoint->bEndpointAddress,
&urbcount);
......
// select the pipe object
pipe = &ar_usb->pipes[pipe_num];
// initialize the ar_usb field
pipe->ar_usb = ar_usb;
}
The driver assumes that the addresses reported in endpoint
descriptors from device side to be complete. If a device is
malicious and does not report complete addresses, it may trigger
NULL-ptr-deref `ath6kl_usb_alloc_urb_from_pipe` and
`ath6kl_usb_free_urb_to_pipe`.
This patch fixes the bug by preventing potential NULL-ptr-deref.
Signed-off-by: Hui Peng <benquike@gmail.com>
Reported-by: Hui Peng <benquike@gmail.com>
Reported-by: Mathias Payer <mathias.payer@nebelwelt.net>
---
drivers/net/wireless/ath/ath6kl/usb.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/net/wireless/ath/ath6kl/usb.c b/drivers/net/wireless/ath/ath6kl/usb.c
index 4defb7a0330f..53b66e9434c9 100644
--- a/drivers/net/wireless/ath/ath6kl/usb.c
+++ b/drivers/net/wireless/ath/ath6kl/usb.c
@@ -132,6 +132,10 @@ ath6kl_usb_alloc_urb_from_pipe(struct ath6kl_usb_pipe *pipe)
struct ath6kl_urb_context *urb_context = NULL;
unsigned long flags;
+ /* bail if this pipe is not initialized */
+ if (!pipe->ar_usb)
+ return NULL;
+
spin_lock_irqsave(&pipe->ar_usb->cs_lock, flags);
if (!list_empty(&pipe->urb_list_head)) {
urb_context =
@@ -150,6 +154,10 @@ static void ath6kl_usb_free_urb_to_pipe(struct ath6kl_usb_pipe *pipe,
{
unsigned long flags;
+ /* bail if this pipe is not initialized */
+ if (!pipe->ar_usb)
+ return;
+
spin_lock_irqsave(&pipe->ar_usb->cs_lock, flags);
pipe->urb_cnt++;
--
2.22.0
^ permalink raw reply related
* [PATCH 2/2] Fix a NULL-ptr-deref bug in ath10k_usb_alloc_urb_from_pipe
From: Hui Peng @ 2019-08-04 0:31 UTC (permalink / raw)
To: kvalo, davem
Cc: Hui Peng, Mathias Payer, ath10k, linux-wireless, netdev,
linux-kernel
The `ar_usb` field of `ath10k_usb_pipe_usb_pipe` objects
are initialized to point to the containing `ath10k_usb` object
according to endpoint descriptors read from the device side, as shown
below in `ath10k_usb_setup_pipe_resources`:
for (i = 0; i < iface_desc->desc.bNumEndpoints; ++i) {
endpoint = &iface_desc->endpoint[i].desc;
// get the address from endpoint descriptor
pipe_num = ath10k_usb_get_logical_pipe_num(ar_usb,
endpoint->bEndpointAddress,
&urbcount);
......
// select the pipe object
pipe = &ar_usb->pipes[pipe_num];
// initialize the ar_usb field
pipe->ar_usb = ar_usb;
}
The driver assumes that the addresses reported in endpoint
descriptors from device side to be complete. If a device is
malicious and does not report complete addresses, it may trigger
NULL-ptr-deref `ath10k_usb_alloc_urb_from_pipe` and
`ath10k_usb_free_urb_to_pipe`.
This patch fixes the bug by preventing potential NULL-ptr-deref.
Signed-off-by: Hui Peng <benquike@gmail.com>
Reported-by: Hui Peng <benquike@gmail.com>
Reported-by: Mathias Payer <mathias.payer@nebelwelt.net>
---
drivers/net/wireless/ath/ath10k/usb.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/net/wireless/ath/ath10k/usb.c b/drivers/net/wireless/ath/ath10k/usb.c
index e1420f67f776..14d86627b47f 100644
--- a/drivers/net/wireless/ath/ath10k/usb.c
+++ b/drivers/net/wireless/ath/ath10k/usb.c
@@ -38,6 +38,10 @@ ath10k_usb_alloc_urb_from_pipe(struct ath10k_usb_pipe *pipe)
struct ath10k_urb_context *urb_context = NULL;
unsigned long flags;
+ /* bail if this pipe is not initialized */
+ if (!pipe->ar_usb)
+ return NULL;
+
spin_lock_irqsave(&pipe->ar_usb->cs_lock, flags);
if (!list_empty(&pipe->urb_list_head)) {
urb_context = list_first_entry(&pipe->urb_list_head,
@@ -55,6 +59,10 @@ static void ath10k_usb_free_urb_to_pipe(struct ath10k_usb_pipe *pipe,
{
unsigned long flags;
+ /* bail if this pipe is not initialized */
+ if (!pipe->ar_usb)
+ return NULL;
+
spin_lock_irqsave(&pipe->ar_usb->cs_lock, flags);
pipe->urb_cnt++;
--
2.22.0
^ permalink raw reply related
* Re: [PATCH net-next] net/tls: prevent skb_orphan() from leaking TLS plain text with offload
From: Jakub Kicinski @ 2019-08-04 1:58 UTC (permalink / raw)
To: David Miller
Cc: netdev, oss-drivers, edumazet, borisp, aviadye, davejwatson,
john.fastabend, daniel, willemb, xiyou.wangcong, fw,
alexei.starovoitov
In-Reply-To: <20190802.172453.1075167824005057182.davem@davemloft.net>
On Fri, 02 Aug 2019 17:24:53 -0700 (PDT), David Miller wrote:
> From: Jakub Kicinski <jakub.kicinski@netronome.com>
> Date: Tue, 30 Jul 2019 14:12:58 -0700
>
> > I'm sending this for net-next because of lack of confidence
> > in my own abilities. It should apply cleanly to net... :)
>
> It looks like there will be changes to this.
Yes, sorry for going silent, I reworked this to follow Boris'es
suggestion of using flags, and wanted a little bit of extra QA.
Unfortunately 5.3-rc has broken Intel IOMMU and QA lab machines
don't boot.. :(
^ permalink raw reply
* Re: [v2,1/2] tools: bpftool: add net attach command to attach XDP on interface
From: Jakub Kicinski @ 2019-08-04 2:13 UTC (permalink / raw)
To: Daniel T. Lee; +Cc: Daniel Borkmann, Alexei Starovoitov, netdev
In-Reply-To: <CAEKGpzg91CMvq5FnYhAxX7XoA+Sr-+AY3t34q5Q2meG93Ydq9Q@mail.gmail.com>
On Sat, 3 Aug 2019 18:39:21 +0900, Daniel T. Lee wrote:
> On Sat, Aug 3, 2019 at 3:39 AM Jakub Kicinski wrote:
> > Right, I was wondering if we want to call it force, though? force is
> > sort of a reuse of iproute2 concept. But it's kind of hard to come up
> > with names.
> >
> > Just to be sure - I mean something like:
> >
> > bpftool net attach xdp id xyz dev ethN noreplace
> >
> > Rather than:
> >
> > bpftool -f net attach xdp id xyz dev ethN
> >
>
> How about a word 'immutable'? 'noreplace' seems good though.
> Just suggesting an option.
Hm. Immutable kind of has a meaning in Linux (check out immutable in
extended file attributes, and CAP_LINUX_IMMUTABLE) which is different
than here.. so I'd avoid that one.
Another option I was thinking about was using the same keywords as maps
do: "noexist" - although we don't have a way of doing "exist"
currently, which is kind of breaking the equivalence.
Or maye we should go the same route as iproute2 after all, and set
noreplace by default? That way we don't need the negation in the
name. We could use "overwrite", "replace"?
Naming things... :)
> > > > > +}
> > > > > +
> > > > > +static int do_attach(int argc, char **argv)
> > > > > +{
> > > > > + enum net_attach_type attach_type;
> > > > > + int err, progfd, ifindex;
> > > > > +
> > > > > + err = parse_attach_args(argc, argv, &progfd, &attach_type, &ifindex);
> > > > > + if (err)
> > > > > + return err;
> > > >
> > > > Probably not the best idea to move this out into a helper.
> > >
> > > Again, just trying to make consistent with 'prog.c'.
> > >
> > > But clearly it has differences with do_attach/detach from 'prog.c'.
> > > From it, it uses the same parse logic 'parse_attach_detach_args' since
> > > the two command 'bpftool prog attach/detach' uses the same argument format.
> > >
> > > However, in here, 'bpftool net' attach and detach requires different number of
> > > argument, so function for parse argument has been defined separately.
> > > The situation is little bit different, but keeping argument parse logic as an
> > > helper, I think it's better in terms of consistency.
> >
> > Well they won't share the same arguments if you add the keyword for
> > controlling IF_NOEXIST :(
>
> My apologies, but I think I'm not following with you.
>
> Currently detach/attach isn't sharing same arguments.
> And even after adding the argument for IF_NOEXIST such as [ noreplace ],
> it'll stays the same for not sharing same arguments.
Ah, my apologies I misread your message.
> I'm not sure why it is not the best idea to move arg parse logic into a helper.
Output args are kind of ugly, and having to pass each parameter through
output arg seems like something that could get out of hand as the
number grows.
Usually command handling functions are relatively small and
straightforward in bpftool so it's quite nice to have it all in one
simple procedure.
But I'm not feeling too strongly about it. Feel free to leave the
parsing in separate functions if you prefer!
^ permalink raw reply
* Re: [PATCH net-next v3] net: phy: broadcom: add 1000Base-X support for BCM54616S
From: Tao Ren @ 2019-08-04 4:48 UTC (permalink / raw)
To: Vladimir Oltean
Cc: Andrew Lunn, Florian Fainelli, Heiner Kallweit, David S . Miller,
Arun Parameswaran, Justin Chen, netdev, lkml,
openbmc@lists.ozlabs.org
In-Reply-To: <CA+h21hrOEape89MTqCUyGFt=f6ba7Q-2KcOsN_Vw2Qv8iq86jw@mail.gmail.com>
Hi Vladimir,
On 8/3/19 6:49 AM, Vladimir Oltean wrote:
> Hi Tao,
>
> On Sat, 3 Aug 2019 at 00:56, Tao Ren <taoren@fb.com> wrote:
>>
>> genphy_read_status() cannot report correct link speed when BCM54616S PHY
>> is configured in RGMII->1000Base-KX mode (for example, on Facebook CMM
>> BMC platform), and it is because speed-related fields in MII registers
>> are assigned different meanings in 1000X register set. Actually there
>> is no speed field in 1000X register set because link speed is always
>> 1000 Mb/s.
>>
>> The patch adds "probe" callback to detect PHY's operation mode based on
>> INTERF_SEL[1:0] pins and 1000X/100FX selection bit in SerDES 100-FX
>> Control register. Besides, link speed is manually set to 1000 Mb/s in
>> "read_status" callback if PHY-switch link is 1000Base-X.
>>
>> Signed-off-by: Tao Ren <taoren@fb.com>
>> ---
>> Changes in v3:
>> - rename bcm5482_read_status to bcm54xx_read_status so the callback can
>> be shared by BCM5482 and BCM54616S.
>> Changes in v2:
>> - Auto-detect PHY operation mode instead of passing DT node.
>> - move PHY mode auto-detect logic from config_init to probe callback.
>> - only set speed (not including duplex) in read_status callback.
>> - update patch description with more background to avoid confusion.
>> - patch #1 in the series ("net: phy: broadcom: set features explicitly
>> for BCM54616") is dropped: the fix should go to get_features callback
>> which may potentially depend on this patch.
>>
>> drivers/net/phy/broadcom.c | 41 +++++++++++++++++++++++++++++++++-----
>> include/linux/brcmphy.h | 10 ++++++++--
>> 2 files changed, 44 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/net/phy/broadcom.c b/drivers/net/phy/broadcom.c
>> index 937d0059e8ac..ecad8a201a09 100644
>> --- a/drivers/net/phy/broadcom.c
>> +++ b/drivers/net/phy/broadcom.c
>> @@ -383,9 +383,9 @@ static int bcm5482_config_init(struct phy_device *phydev)
>> /*
>> * Select 1000BASE-X register set (primary SerDes)
>> */
>> - reg = bcm_phy_read_shadow(phydev, BCM5482_SHD_MODE);
>> - bcm_phy_write_shadow(phydev, BCM5482_SHD_MODE,
>> - reg | BCM5482_SHD_MODE_1000BX);
>> + reg = bcm_phy_read_shadow(phydev, BCM54XX_SHD_MODE);
>> + bcm_phy_write_shadow(phydev, BCM54XX_SHD_MODE,
>> + reg | BCM54XX_SHD_MODE_1000BX);
>>
>> /*
>> * LED1=ACTIVITYLED, LED3=LINKSPD[2]
>> @@ -409,7 +409,7 @@ static int bcm5482_config_init(struct phy_device *phydev)
>> return err;
>> }
>>
>> -static int bcm5482_read_status(struct phy_device *phydev)
>> +static int bcm54xx_read_status(struct phy_device *phydev)
>> {
>> int err;
>>
>> @@ -464,6 +464,35 @@ static int bcm54616s_config_aneg(struct phy_device *phydev)
>> return ret;
>> }
>>
>> +static int bcm54616s_probe(struct phy_device *phydev)
>> +{
>> + int val, intf_sel;
>> +
>> + val = bcm_phy_read_shadow(phydev, BCM54XX_SHD_MODE);
>> + if (val < 0)
>> + return val;
>> +
>> + /* The PHY is strapped in RGMII to fiber mode when INTERF_SEL[1:0]
>> + * is 01b.
>> + */
>> + intf_sel = (val & BCM54XX_SHD_INTF_SEL_MASK) >> 1;
>> + if (intf_sel == 1) {
>> + val = bcm_phy_read_shadow(phydev, BCM54616S_SHD_100FX_CTRL);
>> + if (val < 0)
>> + return val;
>> +
>> + /* Bit 0 of the SerDes 100-FX Control register, when set
>> + * to 1, sets the MII/RGMII -> 100BASE-FX configuration.
>> + * When this bit is set to 0, it sets the GMII/RGMII ->
>> + * 1000BASE-X configuration.
>> + */
>> + if (!(val & BCM54616S_100FX_MODE))
>> + phydev->dev_flags |= PHY_BCM_FLAGS_MODE_1000BX;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> static int brcm_phy_setbits(struct phy_device *phydev, int reg, int set)
>> {
>> int val;
>> @@ -655,6 +684,8 @@ static struct phy_driver broadcom_drivers[] = {
>> .config_aneg = bcm54616s_config_aneg,
>> .ack_interrupt = bcm_phy_ack_intr,
>> .config_intr = bcm_phy_config_intr,
>> + .read_status = bcm54xx_read_status,
>> + .probe = bcm54616s_probe,
>> }, {
>> .phy_id = PHY_ID_BCM5464,
>> .phy_id_mask = 0xfffffff0,
>> @@ -689,7 +720,7 @@ static struct phy_driver broadcom_drivers[] = {
>> .name = "Broadcom BCM5482",
>> /* PHY_GBIT_FEATURES */
>> .config_init = bcm5482_config_init,
>> - .read_status = bcm5482_read_status,
>> + .read_status = bcm54xx_read_status,
>> .ack_interrupt = bcm_phy_ack_intr,
>> .config_intr = bcm_phy_config_intr,
>> }, {
>> diff --git a/include/linux/brcmphy.h b/include/linux/brcmphy.h
>> index 6db2d9a6e503..b475e7f20d28 100644
>> --- a/include/linux/brcmphy.h
>> +++ b/include/linux/brcmphy.h
>> @@ -200,9 +200,15 @@
>> #define BCM5482_SHD_SSD 0x14 /* 10100: Secondary SerDes control */
>> #define BCM5482_SHD_SSD_LEDM 0x0008 /* SSD LED Mode enable */
>> #define BCM5482_SHD_SSD_EN 0x0001 /* SSD enable */
>> -#define BCM5482_SHD_MODE 0x1f /* 11111: Mode Control Register */
>> -#define BCM5482_SHD_MODE_1000BX 0x0001 /* Enable 1000BASE-X registers */
>>
>> +/* 10011: SerDes 100-FX Control Register */
>> +#define BCM54616S_SHD_100FX_CTRL 0x13
>> +#define BCM54616S_100FX_MODE BIT(0) /* 100-FX SerDes Enable */
>> +
>> +/* 11111: Mode Control Register */
>> +#define BCM54XX_SHD_MODE 0x1f
>> +#define BCM54XX_SHD_INTF_SEL_MASK GENMASK(2, 1) /* INTERF_SEL[1:0] */
>> +#define BCM54XX_SHD_MODE_1000BX BIT(0) /* Enable 1000-X registers */
>>
>> /*
>> * EXPANSION SHADOW ACCESS REGISTERS. (PHY REG 0x15, 0x16, and 0x17)
>> --
>> 2.17.1
>>
>
> The patchset looks better now. But is it ok, I wonder, to keep
> PHY_BCM_FLAGS_MODE_1000BX in phydev->dev_flags, considering that
> phy_attach_direct is overwriting it?
I checked ftgmac100 driver (used on my machine) and it calls phy_connect_direct which passes phydev->dev_flags when calling phy_attach_direct: that explains why the flag is not cleared in my case.
Can you give me some suggestions since dev_flags may be override in other calling paths? For example, is it good idea to move the auto-detect logic from probe to config_init? Or are there other fields in phy_device structure that can be used to store PHY-switch link type? Or maybe I should just include the auto-detect logic in read_status callback?
Thanks,
Tao
^ permalink raw reply
* Re: [PATCH bpf-next 1/2] selftests/bpf: add loop test 4
From: Yonghong Song @ 2019-08-04 5:29 UTC (permalink / raw)
To: Alexei Starovoitov, davem@davemloft.net
Cc: daniel@iogearbox.net, netdev@vger.kernel.org, bpf@vger.kernel.org,
Kernel Team
In-Reply-To: <20190802233344.863418-2-ast@kernel.org>
On 8/2/19 4:33 PM, Alexei Starovoitov wrote:
> Add a test that returns a 'random' number between [0, 2^20)
> If state pruning is not working correctly for loop body the number of
> processed insns will be 2^20 * num_of_insns_in_loop_body and the program
> will be rejected.
The maximum processed insns will be 2^20 or 2^20 *
num_of_insns_in_loop_body? I thought the verifier will
stop processing once processed insns reach 1M?
Could you elaborate which potential issues in verifier
you try to cover with this test case? Extra tests are
always welcome. We already have scale/loop tests and some
(e.g., strobemeta tests) are more complex than this one.
Maybe you have something in mind for this particular
test? Putting in the commit message may help people understand
the concerns.
>
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
> ---
> .../bpf/prog_tests/bpf_verif_scale.c | 1 +
> tools/testing/selftests/bpf/progs/loop4.c | 23 +++++++++++++++++++
> 2 files changed, 24 insertions(+)
> create mode 100644 tools/testing/selftests/bpf/progs/loop4.c
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c b/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
> index b4be96162ff4..757e39540eda 100644
> --- a/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
> +++ b/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
> @@ -71,6 +71,7 @@ void test_bpf_verif_scale(void)
>
> { "loop1.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
> { "loop2.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
> + { "loop4.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
The program is more like a BPF_PROG_TYPE_SCHED_CLS type than
a BPF_PROG_TYPE_RAW_TRACEPOINT?
>
> /* partial unroll. 19k insn in a loop.
> * Total program size 20.8k insn.
> diff --git a/tools/testing/selftests/bpf/progs/loop4.c b/tools/testing/selftests/bpf/progs/loop4.c
> new file mode 100644
> index 000000000000..3e7ee14fddbd
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/loop4.c
> @@ -0,0 +1,23 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Copyright (c) 2019 Facebook
> +#include <linux/sched.h>
> +#include <linux/ptrace.h>
Since the program is a networking type,
the above two headers are probably unneeded.
> +#include <stdint.h>
> +#include <stddef.h>
> +#include <stdbool.h>
> +#include <linux/bpf.h>
> +#include "bpf_helpers.h"
> +
> +char _license[] SEC("license") = "GPL";
> +
> +SEC("socket")
> +int combinations(volatile struct __sk_buff* skb)
> +{
> + int ret = 0, i;
> +
> +#pragma nounroll
> + for (i = 0; i < 20; i++)
> + if (skb->len)
> + ret |= 1 << i;
> + return ret;
> +}
>
^ permalink raw reply
* Re: [PATCH bpf-next 2/2] selftests/bpf: add loop test 5
From: Yonghong Song @ 2019-08-04 5:45 UTC (permalink / raw)
To: Alexei Starovoitov, davem@davemloft.net
Cc: daniel@iogearbox.net, netdev@vger.kernel.org, bpf@vger.kernel.org,
Kernel Team
In-Reply-To: <20190802233344.863418-3-ast@kernel.org>
On 8/2/19 4:33 PM, Alexei Starovoitov wrote:
> Add a test with multiple exit conditions.
> It's not an infinite loop only when the verifier can properly track
> all math on variable 'i' through all possible ways of executing this loop.
Agreed with motivation of this test.
>
> barrier()s are needed to disable llvm optimization that combines multiple
> branches into fewer branches.
>
> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
> ---
> .../bpf/prog_tests/bpf_verif_scale.c | 1 +
> tools/testing/selftests/bpf/progs/loop5.c | 37 +++++++++++++++++++
> 2 files changed, 38 insertions(+)
> create mode 100644 tools/testing/selftests/bpf/progs/loop5.c
>
> diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c b/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
> index 757e39540eda..29615a4a9362 100644
> --- a/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
> +++ b/tools/testing/selftests/bpf/prog_tests/bpf_verif_scale.c
> @@ -72,6 +72,7 @@ void test_bpf_verif_scale(void)
> { "loop1.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
> { "loop2.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
> { "loop4.o", BPF_PROG_TYPE_RAW_TRACEPOINT }, > + { "loop5.o", BPF_PROG_TYPE_RAW_TRACEPOINT },
More like a BPF_PROG_TYPE_SCHED_CLS type although probably it does not
matter as we did not attach it to anywhere?
>
> /* partial unroll. 19k insn in a loop.
> * Total program size 20.8k insn.
> diff --git a/tools/testing/selftests/bpf/progs/loop5.c b/tools/testing/selftests/bpf/progs/loop5.c
> new file mode 100644
> index 000000000000..9d9817efe208
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/progs/loop5.c
> @@ -0,0 +1,37 @@
> +// SPDX-License-Identifier: GPL-2.0
> +// Copyright (c) 2019 Facebook
> +#include <linux/sched.h>
> +#include <linux/ptrace.h>
The above headers probably not needed.
> +#include <stdint.h>
> +#include <stddef.h>
> +#include <stdbool.h>
> +#include <linux/bpf.h>
> +#include "bpf_helpers.h"
> +#define barrier() __asm__ __volatile__("": : :"memory")
> +
> +char _license[] SEC("license") = "GPL";
> +
> +SEC("socket")
> +int while_true(volatile struct __sk_buff* skb)
> +{
> + int i = 0;
> +
> + while (true) {
> + if (skb->len)
> + i += 3;
> + else
> + i += 7;
> + if (i == 9)
> + break;
> + barrier();
> + if (i == 10)
> + break;
> + barrier();
> + if (i == 13)
> + break;
> + barrier();
> + if (i == 14)
> + break;
> + }
> + return i;
> +}
>
^ permalink raw reply
* Re: [PATCH 1/1] bpf: introduce new helper udp_flow_src_port
From: Y Song @ 2019-08-04 6:52 UTC (permalink / raw)
To: Farid Zakaria; +Cc: Alexei Starovoitov, Daniel Borkmann, netdev, bpf
In-Reply-To: <20190803044320.5530-2-farid.m.zakaria@gmail.com>
On Sat, Aug 3, 2019 at 8:29 PM Farid Zakaria <farid.m.zakaria@gmail.com> wrote:
>
> Foo over UDP uses UDP encapsulation to add additional entropy
> into the packets so that they get beter distribution across EMCP
> routes.
>
> Expose udp_flow_src_port as a bpf helper so that tunnel filters
> can benefit from the helper.
>
> Signed-off-by: Farid Zakaria <farid.m.zakaria@gmail.com>
> ---
> include/uapi/linux/bpf.h | 21 +++++++--
> net/core/filter.c | 20 ++++++++
> tools/include/uapi/linux/bpf.h | 21 +++++++--
> tools/testing/selftests/bpf/bpf_helpers.h | 2 +
> .../bpf/prog_tests/udp_flow_src_port.c | 28 +++++++++++
> .../bpf/progs/test_udp_flow_src_port_kern.c | 47 +++++++++++++++++++
> 6 files changed, 131 insertions(+), 8 deletions(-)
> create mode 100644 tools/testing/selftests/bpf/prog_tests/udp_flow_src_port.c
> create mode 100644 tools/testing/selftests/bpf/progs/test_udp_flow_src_port_kern.c
First, for each review, backport and sync with libbpf repo, in the future,
could you break the patch to two patches?
1. kernel changes (net/core/filter.c, include/uapi/linux/bpf.h)
2. tools/include/uapi/linux/bpf.h
3. tools/testing/ changes
Second, could you explain why existing __sk_buff->hash not enough?
there are corner cases where if __sk_buff->hash is 0 and the kernel did some
additional hashing, but maybe you can approximate in bpf program?
For case, min >= max, I suppose you can get min/max port values
from the user space for a particular net device and then calculate
the hash in the bpf program?
What I want to know if how much accuracy you will lose if you just
use __sk_buff->hash and do approximation in bpf program.
>
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index 4393bd4b2419..90e814153dec 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -2545,9 +2545,21 @@ union bpf_attr {
> * *th* points to the start of the TCP header, while *th_len*
> * contains **sizeof**\ (**struct tcphdr**).
> *
> - * Return
> - * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
> - * error otherwise.
> + * Return
> + * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative
> + * error otherwise.
> + *
> + * int bpf_udp_flow_src_port(struct sk_buff *skb, int min, int max, int use_eth)
> + * Description
> + * It's common to implement tunnelling inside a UDP protocol to provide
> + * additional randomness to the packet. The destination port of the UDP
> + * header indicates the inner packet type whereas the source port is used
> + * for additional entropy.
> + *
> + * Return
> + * An obfuscated hash of the packet that falls within the
> + * min & max port range.
> + * If min >= max, the default port range is used
> *
> * int bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags)
> * Description
> @@ -2853,7 +2865,8 @@ union bpf_attr {
> FN(sk_storage_get), \
> FN(sk_storage_delete), \
> FN(send_signal), \
> - FN(tcp_gen_syncookie),
> + FN(tcp_gen_syncookie), \
> + FN(udp_flow_src_port),
>
> /* integer value in 'imm' field of BPF_CALL instruction selects which helper
> * function eBPF program intends to call
> diff --git a/net/core/filter.c b/net/core/filter.c
> index 5a2707918629..fdf0ebb8c2c8 100644
> --- a/net/core/filter.c
> +++ b/net/core/filter.c
> @@ -2341,6 +2341,24 @@ static const struct bpf_func_proto bpf_msg_pull_data_proto = {
> .arg4_type = ARG_ANYTHING,
> };
>
> +BPF_CALL_4(bpf_udp_flow_src_port, struct sk_buff *, skb, int, min,
> + int, max, int, use_eth)
> +{
> + struct net *net = dev_net(skb->dev);
> +
> + return udp_flow_src_port(net, skb, min, max, use_eth);
> +}
> +
[...]
^ permalink raw reply
* Re: [PATCH v2] net/mlx5e: always initialize frag->last_in_page
From: Tariq Toukan @ 2019-08-04 7:45 UTC (permalink / raw)
To: Qian Cai, davem@davemloft.net
Cc: Saeed Mahameed, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org
In-Reply-To: <1564667574-31542-1-git-send-email-cai@lca.pw>
On 8/1/2019 4:52 PM, Qian Cai wrote:
> The commit 069d11465a80 ("net/mlx5e: RX, Enhance legacy Receive Queue
> memory scheme") introduced an undefined behaviour below due to
> "frag->last_in_page" is only initialized in mlx5e_init_frags_partition()
> when,
>
> if (next_frag.offset + frag_info[f].frag_stride > PAGE_SIZE)
>
> or after bailed out the loop,
>
> for (i = 0; i < mlx5_wq_cyc_get_size(&rq->wqe.wq); i++)
>
> As the result, there could be some "frag" have uninitialized
> value of "last_in_page".
>
> Later, get_frag() obtains those "frag" and check "frag->last_in_page" in
> mlx5e_put_rx_frag() and triggers the error during boot. Fix it by always
> initializing "frag->last_in_page" to "false" in
> mlx5e_init_frags_partition().
>
> UBSAN: Undefined behaviour in
> drivers/net/ethernet/mellanox/mlx5/core/en_rx.c:325:12
> load of value 170 is not a valid value for type 'bool' (aka '_Bool')
> Call trace:
> dump_backtrace+0x0/0x264
> show_stack+0x20/0x2c
> dump_stack+0xb0/0x104
> __ubsan_handle_load_invalid_value+0x104/0x128
> mlx5e_handle_rx_cqe+0x8e8/0x12cc [mlx5_core]
> mlx5e_poll_rx_cq+0xca8/0x1a94 [mlx5_core]
> mlx5e_napi_poll+0x17c/0xa30 [mlx5_core]
> net_rx_action+0x248/0x940
> __do_softirq+0x350/0x7b8
> irq_exit+0x200/0x26c
> __handle_domain_irq+0xc8/0x128
> gic_handle_irq+0x138/0x228
> el1_irq+0xb8/0x140
> arch_cpu_idle+0x1a4/0x348
> do_idle+0x114/0x1b0
> cpu_startup_entry+0x24/0x28
> rest_init+0x1ac/0x1dc
> arch_call_rest_init+0x10/0x18
> start_kernel+0x4d4/0x57c
>
> Fixes: 069d11465a80 ("net/mlx5e: RX, Enhance legacy Receive Queue memory scheme")
> Signed-off-by: Qian Cai <cai@lca.pw>
> ---
>
> v2: zero-init the whole struct instead per Tariq.
>
> drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 5 ++---
> 1 file changed, 2 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
> index 47eea6b3a1c3..e1810c03a510 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
> @@ -331,12 +331,11 @@ static inline u64 mlx5e_get_mpwqe_offset(struct mlx5e_rq *rq, u16 wqe_ix)
>
> static void mlx5e_init_frags_partition(struct mlx5e_rq *rq)
> {
> - struct mlx5e_wqe_frag_info next_frag, *prev;
> + struct mlx5e_wqe_frag_info next_frag = {};
> + struct mlx5e_wqe_frag_info *prev = NULL;
> int i;
>
> next_frag.di = &rq->wqe.di[0];
> - next_frag.offset = 0;
> - prev = NULL;
>
> for (i = 0; i < mlx5_wq_cyc_get_size(&rq->wqe.wq); i++) {
> struct mlx5e_rq_frag_info *frag_info = &rq->wqe.info.arr[0];
>
Reviewed-by: Tariq Toukan <tariqt@mellanox.com>
Thanks.
^ permalink raw reply
* [PATCH net-next] r8169: remove access to legacy register MultiIntr
From: Heiner Kallweit @ 2019-08-04 7:42 UTC (permalink / raw)
To: Realtek linux nic maintainers, David Miller; +Cc: netdev@vger.kernel.org
This code piece was inherited from RTL8139 code, the register at
address 0x5c however has a different meaning on RTL8169 and is unused.
So we can remove this.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
drivers/net/ethernet/realtek/r8169_main.c | 4 ----
1 file changed, 4 deletions(-)
diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 0be8e5c08..e38bc01eb 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -271,7 +271,6 @@ enum rtl_registers {
Config3 = 0x54,
Config4 = 0x55,
Config5 = 0x56,
- MultiIntr = 0x5c,
PHYAR = 0x60,
PHYstatus = 0x6c,
RxMaxSize = 0xda,
@@ -5241,10 +5240,7 @@ static void rtl_hw_start(struct rtl8169_private *tp)
RTL_W8(tp, ChipCmd, CmdTxEnb | CmdRxEnb);
rtl_init_rxcfg(tp);
rtl_set_tx_config_registers(tp);
-
rtl_set_rx_mode(tp->dev);
- /* no early-rx interrupts */
- RTL_W16(tp, MultiIntr, RTL_R16(tp, MultiIntr) & 0xf000);
rtl_irq_enable(tp);
}
--
2.22.0
^ permalink raw reply related
* [PATCH net-next] r8169: add helper r8168_mac_ocp_modify
From: Heiner Kallweit @ 2019-08-04 7:47 UTC (permalink / raw)
To: Realtek linux nic maintainers, David Miller; +Cc: netdev@vger.kernel.org
Add a helper for MAC OCP read-modify-write operations.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
drivers/net/ethernet/realtek/r8169_main.c | 65 +++++++----------------
1 file changed, 19 insertions(+), 46 deletions(-)
diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index e38bc01eb..039a967c7 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -850,6 +850,14 @@ static u16 r8168_mac_ocp_read(struct rtl8169_private *tp, u32 reg)
return RTL_R32(tp, OCPDR);
}
+static void r8168_mac_ocp_modify(struct rtl8169_private *tp, u32 reg, u16 mask,
+ u16 set)
+{
+ u16 data = r8168_mac_ocp_read(tp, reg);
+
+ r8168_mac_ocp_write(tp, reg, (data & ~mask) | set);
+}
+
#define OCP_STD_PHY_BASE 0xa400
static void r8168g_mdio_write(struct rtl8169_private *tp, int reg, int value)
@@ -4809,8 +4817,6 @@ static void rtl_hw_start_8411_2(struct rtl8169_private *tp)
static void rtl_hw_start_8168h_1(struct rtl8169_private *tp)
{
- int rg_saw_cnt;
- u32 data;
static const struct ephy_info e_info_8168h_1[] = {
{ 0x1e, 0x0800, 0x0001 },
{ 0x1d, 0x0000, 0x0800 },
@@ -4819,6 +4825,7 @@ static void rtl_hw_start_8168h_1(struct rtl8169_private *tp)
{ 0x04, 0xffff, 0x154a },
{ 0x01, 0xffff, 0x068b }
};
+ int rg_saw_cnt;
/* disable aspm and clock request before access ephy */
rtl_hw_aspm_clkreq_enable(tp, false);
@@ -4863,31 +4870,13 @@ static void rtl_hw_start_8168h_1(struct rtl8169_private *tp)
sw_cnt_1ms_ini = 16000000/rg_saw_cnt;
sw_cnt_1ms_ini &= 0x0fff;
- data = r8168_mac_ocp_read(tp, 0xd412);
- data &= ~0x0fff;
- data |= sw_cnt_1ms_ini;
- r8168_mac_ocp_write(tp, 0xd412, data);
+ r8168_mac_ocp_modify(tp, 0xd412, 0x0fff, sw_cnt_1ms_ini);
}
- data = r8168_mac_ocp_read(tp, 0xe056);
- data &= ~0xf0;
- data |= 0x70;
- r8168_mac_ocp_write(tp, 0xe056, data);
-
- data = r8168_mac_ocp_read(tp, 0xe052);
- data &= ~0x6000;
- data |= 0x8008;
- r8168_mac_ocp_write(tp, 0xe052, data);
-
- data = r8168_mac_ocp_read(tp, 0xe0d6);
- data &= ~0x01ff;
- data |= 0x017f;
- r8168_mac_ocp_write(tp, 0xe0d6, data);
-
- data = r8168_mac_ocp_read(tp, 0xd420);
- data &= ~0x0fff;
- data |= 0x047f;
- r8168_mac_ocp_write(tp, 0xd420, data);
+ r8168_mac_ocp_modify(tp, 0xe056, 0x00f0, 0x0070);
+ r8168_mac_ocp_modify(tp, 0xe052, 0x6000, 0x8008);
+ r8168_mac_ocp_modify(tp, 0xe0d6, 0x01ff, 0x017f);
+ r8168_mac_ocp_modify(tp, 0xd420, 0x0fff, 0x047f);
r8168_mac_ocp_write(tp, 0xe63e, 0x0001);
r8168_mac_ocp_write(tp, 0xe63e, 0x0000);
@@ -4969,7 +4958,6 @@ static void rtl_hw_start_8168ep_2(struct rtl8169_private *tp)
static void rtl_hw_start_8168ep_3(struct rtl8169_private *tp)
{
- u32 data;
static const struct ephy_info e_info_8168ep_3[] = {
{ 0x00, 0xffff, 0x10a3 },
{ 0x19, 0xffff, 0x7c00 },
@@ -4986,18 +4974,9 @@ static void rtl_hw_start_8168ep_3(struct rtl8169_private *tp)
RTL_W8(tp, DLLPR, RTL_R8(tp, DLLPR) & ~PFM_EN);
RTL_W8(tp, MISC_1, RTL_R8(tp, MISC_1) & ~PFM_D3COLD_EN);
- data = r8168_mac_ocp_read(tp, 0xd3e2);
- data &= 0xf000;
- data |= 0x0271;
- r8168_mac_ocp_write(tp, 0xd3e2, data);
-
- data = r8168_mac_ocp_read(tp, 0xd3e4);
- data &= 0xff00;
- r8168_mac_ocp_write(tp, 0xd3e4, data);
-
- data = r8168_mac_ocp_read(tp, 0xe860);
- data |= 0x0080;
- r8168_mac_ocp_write(tp, 0xe860, data);
+ r8168_mac_ocp_modify(tp, 0xd3e2, 0x0fff, 0x0271);
+ r8168_mac_ocp_modify(tp, 0xd3e4, 0x00ff, 0x0000);
+ r8168_mac_ocp_modify(tp, 0xe860, 0x0000, 0x0080);
rtl_hw_aspm_clkreq_enable(tp, true);
}
@@ -6659,8 +6638,6 @@ static int r8169_mdio_register(struct rtl8169_private *tp)
static void rtl_hw_init_8168g(struct rtl8169_private *tp)
{
- u32 data;
-
tp->ocp_base = OCP_STD_PHY_BASE;
RTL_W32(tp, MISC, RTL_R32(tp, MISC) | RXDV_GATED_EN);
@@ -6675,16 +6652,12 @@ static void rtl_hw_init_8168g(struct rtl8169_private *tp)
msleep(1);
RTL_W8(tp, MCU, RTL_R8(tp, MCU) & ~NOW_IS_OOB);
- data = r8168_mac_ocp_read(tp, 0xe8de);
- data &= ~(1 << 14);
- r8168_mac_ocp_write(tp, 0xe8de, data);
+ r8168_mac_ocp_modify(tp, 0xe8de, BIT(14), 0);
if (!rtl_udelay_loop_wait_high(tp, &rtl_link_list_ready_cond, 100, 42))
return;
- data = r8168_mac_ocp_read(tp, 0xe8de);
- data |= (1 << 15);
- r8168_mac_ocp_write(tp, 0xe8de, data);
+ r8168_mac_ocp_modify(tp, 0xe8de, 0, BIT(15));
rtl_udelay_loop_wait_high(tp, &rtl_link_list_ready_cond, 100, 42);
}
--
2.22.0
^ permalink raw reply related
* [PATCH net-next] r8169: sync PCIe PHY init with vendor driver 8.047.01
From: Heiner Kallweit @ 2019-08-04 7:52 UTC (permalink / raw)
To: Realtek linux nic maintainers, David Miller; +Cc: netdev@vger.kernel.org
Synchronize PCIe PHY initialization with vendor driver version 8.047.01.
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
drivers/net/ethernet/realtek/r8169_main.c | 60 ++++++++++++++---------
1 file changed, 38 insertions(+), 22 deletions(-)
diff --git a/drivers/net/ethernet/realtek/r8169_main.c b/drivers/net/ethernet/realtek/r8169_main.c
index 039a967c7..3c7af6669 100644
--- a/drivers/net/ethernet/realtek/r8169_main.c
+++ b/drivers/net/ethernet/realtek/r8169_main.c
@@ -4415,7 +4415,7 @@ static void rtl_hw_start_8168c_2(struct rtl8169_private *tp)
{
static const struct ephy_info e_info_8168c_2[] = {
{ 0x01, 0, 0x0001 },
- { 0x03, 0x0400, 0x0220 }
+ { 0x03, 0x0400, 0x0020 }
};
rtl_set_def_aspm_entry_latency(tp);
@@ -4462,7 +4462,8 @@ static void rtl_hw_start_8168d_4(struct rtl8169_private *tp)
static const struct ephy_info e_info_8168d_4[] = {
{ 0x0b, 0x0000, 0x0048 },
{ 0x19, 0x0020, 0x0050 },
- { 0x0c, 0x0100, 0x0020 }
+ { 0x0c, 0x0100, 0x0020 },
+ { 0x10, 0x0004, 0x0000 },
};
rtl_set_def_aspm_entry_latency(tp);
@@ -4512,7 +4513,9 @@ static void rtl_hw_start_8168e_2(struct rtl8169_private *tp)
{
static const struct ephy_info e_info_8168e_2[] = {
{ 0x09, 0x0000, 0x0080 },
- { 0x19, 0x0000, 0x0224 }
+ { 0x19, 0x0000, 0x0224 },
+ { 0x00, 0x0000, 0x0004 },
+ { 0x0c, 0x3df0, 0x0200 },
};
rtl_set_def_aspm_entry_latency(tp);
@@ -4574,7 +4577,9 @@ static void rtl_hw_start_8168f_1(struct rtl8169_private *tp)
{ 0x06, 0x00c0, 0x0020 },
{ 0x08, 0x0001, 0x0002 },
{ 0x09, 0x0000, 0x0080 },
- { 0x19, 0x0000, 0x0224 }
+ { 0x19, 0x0000, 0x0224 },
+ { 0x00, 0x0000, 0x0004 },
+ { 0x0c, 0x3df0, 0x0200 },
};
rtl_hw_start_8168f(tp);
@@ -4589,8 +4594,9 @@ static void rtl_hw_start_8411(struct rtl8169_private *tp)
static const struct ephy_info e_info_8168f_1[] = {
{ 0x06, 0x00c0, 0x0020 },
{ 0x0f, 0xffff, 0x5200 },
- { 0x1e, 0x0000, 0x4000 },
- { 0x19, 0x0000, 0x0224 }
+ { 0x19, 0x0000, 0x0224 },
+ { 0x00, 0x0000, 0x0004 },
+ { 0x0c, 0x3df0, 0x0200 },
};
rtl_hw_start_8168f(tp);
@@ -4629,8 +4635,8 @@ static void rtl_hw_start_8168g(struct rtl8169_private *tp)
static void rtl_hw_start_8168g_1(struct rtl8169_private *tp)
{
static const struct ephy_info e_info_8168g_1[] = {
- { 0x00, 0x0000, 0x0008 },
- { 0x0c, 0x37d0, 0x0820 },
+ { 0x00, 0x0008, 0x0000 },
+ { 0x0c, 0x3ff0, 0x0820 },
{ 0x1e, 0x0000, 0x0001 },
{ 0x19, 0x8000, 0x0000 }
};
@@ -4646,10 +4652,15 @@ static void rtl_hw_start_8168g_1(struct rtl8169_private *tp)
static void rtl_hw_start_8168g_2(struct rtl8169_private *tp)
{
static const struct ephy_info e_info_8168g_2[] = {
- { 0x00, 0x0000, 0x0008 },
- { 0x0c, 0x3df0, 0x0200 },
- { 0x19, 0xffff, 0xfc00 },
- { 0x1e, 0xffff, 0x20eb }
+ { 0x00, 0x0008, 0x0000 },
+ { 0x0c, 0x3ff0, 0x0820 },
+ { 0x19, 0xffff, 0x7c00 },
+ { 0x1e, 0xffff, 0x20eb },
+ { 0x0d, 0xffff, 0x1666 },
+ { 0x00, 0xffff, 0x10a3 },
+ { 0x06, 0xffff, 0xf050 },
+ { 0x04, 0x0000, 0x0010 },
+ { 0x1d, 0x4000, 0x0000 },
};
rtl_hw_start_8168g(tp);
@@ -4663,11 +4674,16 @@ static void rtl_hw_start_8168g_2(struct rtl8169_private *tp)
static void rtl_hw_start_8411_2(struct rtl8169_private *tp)
{
static const struct ephy_info e_info_8411_2[] = {
- { 0x00, 0x0000, 0x0008 },
- { 0x0c, 0x3df0, 0x0200 },
- { 0x0f, 0xffff, 0x5200 },
- { 0x19, 0x0020, 0x0000 },
- { 0x1e, 0x0000, 0x2000 }
+ { 0x00, 0x0008, 0x0000 },
+ { 0x0c, 0x37d0, 0x0820 },
+ { 0x1e, 0x0000, 0x0001 },
+ { 0x19, 0x8021, 0x0000 },
+ { 0x1e, 0x0000, 0x2000 },
+ { 0x0d, 0x0100, 0x0200 },
+ { 0x00, 0x0000, 0x0080 },
+ { 0x06, 0x0000, 0x0010 },
+ { 0x04, 0x0000, 0x0010 },
+ { 0x1d, 0x0000, 0x4000 },
};
rtl_hw_start_8168g(tp);
@@ -4822,7 +4838,7 @@ static void rtl_hw_start_8168h_1(struct rtl8169_private *tp)
{ 0x1d, 0x0000, 0x0800 },
{ 0x05, 0xffff, 0x2089 },
{ 0x06, 0xffff, 0x5881 },
- { 0x04, 0xffff, 0x154a },
+ { 0x04, 0xffff, 0x854a },
{ 0x01, 0xffff, 0x068b }
};
int rg_saw_cnt;
@@ -4959,10 +4975,10 @@ static void rtl_hw_start_8168ep_2(struct rtl8169_private *tp)
static void rtl_hw_start_8168ep_3(struct rtl8169_private *tp)
{
static const struct ephy_info e_info_8168ep_3[] = {
- { 0x00, 0xffff, 0x10a3 },
- { 0x19, 0xffff, 0x7c00 },
- { 0x1e, 0xffff, 0x20eb },
- { 0x0d, 0xffff, 0x1666 }
+ { 0x00, 0x0000, 0x0080 },
+ { 0x0d, 0x0100, 0x0200 },
+ { 0x19, 0x8021, 0x0000 },
+ { 0x1e, 0x0000, 0x2000 },
};
/* disable aspm and clock request before access ephy */
--
2.22.0
^ permalink raw reply related
* Re: [PATCH V2 7/9] vhost: do not use RCU to synchronize MMU notifier with worker
From: Michael S. Tsirkin @ 2019-08-04 8:07 UTC (permalink / raw)
To: Jason Gunthorpe
Cc: Jason Wang, kvm, virtualization, netdev, linux-kernel, linux-mm
In-Reply-To: <20190804001400.GA25543@ziepe.ca>
On Sat, Aug 03, 2019 at 09:14:00PM -0300, Jason Gunthorpe wrote:
> On Sat, Aug 03, 2019 at 05:36:13PM -0400, Michael S. Tsirkin wrote:
> > On Fri, Aug 02, 2019 at 02:24:18PM -0300, Jason Gunthorpe wrote:
> > > On Fri, Aug 02, 2019 at 10:27:21AM -0400, Michael S. Tsirkin wrote:
> > > > On Fri, Aug 02, 2019 at 09:46:13AM -0300, Jason Gunthorpe wrote:
> > > > > On Fri, Aug 02, 2019 at 05:40:07PM +0800, Jason Wang wrote:
> > > > > > > This must be a proper barrier, like a spinlock, mutex, or
> > > > > > > synchronize_rcu.
> > > > > >
> > > > > >
> > > > > > I start with synchronize_rcu() but both you and Michael raise some
> > > > > > concern.
> > > > >
> > > > > I've also idly wondered if calling synchronize_rcu() under the various
> > > > > mm locks is a deadlock situation.
> > > > >
> > > > > > Then I try spinlock and mutex:
> > > > > >
> > > > > > 1) spinlock: add lots of overhead on datapath, this leads 0 performance
> > > > > > improvement.
> > > > >
> > > > > I think the topic here is correctness not performance improvement
> > > >
> > > > The topic is whether we should revert
> > > > commit 7f466032dc9 ("vhost: access vq metadata through kernel virtual address")
> > > >
> > > > or keep it in. The only reason to keep it is performance.
> > >
> > > Yikes, I'm not sure you can ever win against copy_from_user using
> > > mmu_notifiers?
> >
> > Ever since copy_from_user started playing with flags (for SMAP) and
> > added speculation barriers there's a chance we can win by accessing
> > memory through the kernel address.
>
> You think copy_to_user will be more expensive than the minimum two
> atomics required to synchronize with another thread?
I frankly don't know. With SMAP you flip flags twice, and with spectre
you flush the pipeline. Is that cheaper or more expensive than an atomic
operation? Testing is the only way to tell.
> > > Also, why can't this just permanently GUP the pages? In fact, where
> > > does it put_page them anyhow? Worrying that 7f466 adds a get_user page
> > > but does not add a put_page??
>
> You didn't answer this.. Why not just use GUP?
>
> Jason
Sorry I misunderstood the question. Permanent GUP breaks lots of
functionality we need such as THP and numa balancing.
release_pages is used instead of put_page.
--
MST
^ permalink raw reply
* [PATCH iproute2-next] rdma: Add driver QP type string
From: Gal Pressman @ 2019-08-04 8:07 UTC (permalink / raw)
To: Stephen Hemminger; +Cc: netdev, linux-rdma, Leon Romanovsky, Gal Pressman
RDMA resource tracker now tracks driver QPs as well, add driver QP type
string to qp_types_to_str function.
Signed-off-by: Gal Pressman <galpress@amazon.com>
---
rdma/res.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/rdma/res.c b/rdma/res.c
index ef863f142eca..97a7b9640185 100644
--- a/rdma/res.c
+++ b/rdma/res.c
@@ -148,9 +148,11 @@ const char *qp_types_to_str(uint8_t idx)
"UC", "UD", "RAW_IPV6",
"RAW_ETHERTYPE",
"UNKNOWN", "RAW_PACKET",
- "XRC_INI", "XRC_TGT" };
+ "XRC_INI", "XRC_TGT",
+ [0xFF] = "DRIVER",
+ };
- if (idx < ARRAY_SIZE(qp_types_str))
+ if (idx < ARRAY_SIZE(qp_types_str) && qp_types_str[idx])
return qp_types_str[idx];
return "UNKNOWN";
}
--
2.22.0
^ permalink raw reply related
* Re: [PATCH v4 3/4] net: phy: realtek: Add helpers for accessing RTL8211E extension pages
From: Heiner Kallweit @ 2019-08-04 8:33 UTC (permalink / raw)
To: Matthias Kaehlcke, David S . Miller, Rob Herring, Mark Rutland,
Andrew Lunn, Florian Fainelli
Cc: netdev, devicetree, linux-kernel, Douglas Anderson
In-Reply-To: <20190801190759.28201-4-mka@chromium.org>
On 01.08.2019 21:07, Matthias Kaehlcke wrote:
> The RTL8211E has extension pages, which can be accessed after
> selecting a page through a custom method. Add a function to
> modify bits in a register of an extension page and a helper for
> selecting an ext page. Use rtl8211e_modify_ext_paged() in
> rtl8211e_config_init() instead of doing things 'manually'.
>
> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> ---
> Changes in v4:
> - don't add constant RTL8211E_EXT_PAGE, it's only used once,
> use a literal instead
> - pass 'oldpage' to phy_restore_page() in rtl8211e_select_ext_page(),
> not 'page'
> - return 'oldpage' in rtl8211e_select_ext_page()
> - use __phy_modify() in rtl8211e_modify_ext_paged() instead of
> reimplementing __phy_modify_changed()
> - in rtl8211e_modify_ext_paged() return directly when
> rtl8211e_select_ext_page() fails
> ---
> drivers/net/phy/realtek.c | 48 +++++++++++++++++++++++++++------------
> 1 file changed, 34 insertions(+), 14 deletions(-)
>
> diff --git a/drivers/net/phy/realtek.c b/drivers/net/phy/realtek.c
> index a669945eb829..e09d3b0da2c7 100644
> --- a/drivers/net/phy/realtek.c
> +++ b/drivers/net/phy/realtek.c
> @@ -53,6 +53,36 @@ static int rtl821x_write_page(struct phy_device *phydev, int page)
> return __phy_write(phydev, RTL821x_PAGE_SELECT, page);
> }
>
> +static int rtl8211e_select_ext_page(struct phy_device *phydev, int page)
The "extended page" mechanism doesn't exist on RTL8211E only. A prefix
rtl821x like in other functions may be better therefore.
> +{
> + int ret, oldpage;
> +
> + oldpage = phy_select_page(phydev, 7);
> + if (oldpage < 0)
> + return oldpage;
> +
> + ret = __phy_write(phydev, RTL821x_EXT_PAGE_SELECT, page);
> + if (ret)
> + return phy_restore_page(phydev, oldpage, ret);
> +
> + return oldpage;
> +}
> +
> +static int rtl8211e_modify_ext_paged(struct phy_device *phydev, int page,
> + u32 regnum, u16 mask, u16 set)
> +{
> + int ret = 0;
> + int oldpage;
> +
> + oldpage = rtl8211e_select_ext_page(phydev, page);
> + if (oldpage < 0)
> + return oldpage;
> +
> + ret = __phy_modify(phydev, regnum, mask, set);
> +
> + return phy_restore_page(phydev, oldpage, ret);
> +}
> +
> static int rtl8201_ack_interrupt(struct phy_device *phydev)
> {
> int err;
> @@ -184,7 +214,7 @@ static int rtl8211f_config_init(struct phy_device *phydev)
>
> static int rtl8211e_config_init(struct phy_device *phydev)
> {
> - int ret = 0, oldpage;
> + int ret;
> u16 val;
>
> /* enable TX/RX delay for rgmii-* modes, and disable them for rgmii. */
> @@ -213,19 +243,9 @@ static int rtl8211e_config_init(struct phy_device *phydev)
> * 2 = RX Delay, 1 = TX Delay, 0 = SELRGV (see original PHY datasheet
> * for details).
> */
> - oldpage = phy_select_page(phydev, 0x7);
> - if (oldpage < 0)
> - goto err_restore_page;
> -
> - ret = __phy_write(phydev, RTL821x_EXT_PAGE_SELECT, 0xa4);
> - if (ret)
> - goto err_restore_page;
> -
> - ret = __phy_modify(phydev, 0x1c, RTL8211E_TX_DELAY | RTL8211E_RX_DELAY,
> - val);
> -
> -err_restore_page:
> - return phy_restore_page(phydev, oldpage, ret);
> + return rtl8211e_modify_ext_paged(phydev, 0xa4, 0x1c,
> + RTL8211E_TX_DELAY | RTL8211E_RX_DELAY,
> + val);
> }
>
> static int rtl8211b_suspend(struct phy_device *phydev)
>
^ permalink raw reply
* Re: [PATCH net-next 1/2] net: phy: broadcom: set features explicitly for BCM54616S
From: Heiner Kallweit @ 2019-08-04 8:40 UTC (permalink / raw)
To: Tao Ren, Andrew Lunn
Cc: Florian Fainelli, David S . Miller, Arun Parameswaran,
Justin Chen, Vladimir Oltean, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org, Andrew Jeffery,
openbmc@lists.ozlabs.org
In-Reply-To: <88f4d709-d9bb-943c-37a9-aeebe8ca0ebc@fb.com>
On 01.08.2019 07:20, Tao Ren wrote:
> On 7/30/19 11:00 PM, Tao Ren wrote:
>> On 7/30/19 10:53 PM, Heiner Kallweit wrote:
>>> On 31.07.2019 02:12, Tao Ren wrote:
>>>> On 7/29/19 11:00 PM, Heiner Kallweit wrote:
>>>>> On 30.07.2019 07:05, Tao Ren wrote:
>>>>>> On 7/29/19 8:35 PM, Andrew Lunn wrote:
>>>>>>> On Mon, Jul 29, 2019 at 05:25:32PM -0700, Tao Ren wrote:
>>>>>>>> BCM54616S feature "PHY_GBIT_FEATURES" was removed by commit dcdecdcfe1fc
>>>>>>>> ("net: phy: switch drivers to use dynamic feature detection"). As dynamic
>>>>>>>> feature detection doesn't work when BCM54616S is working in RGMII-Fiber
>>>>>>>> mode (different sets of MII Control/Status registers being used), let's
>>>>>>>> set "PHY_GBIT_FEATURES" for BCM54616S explicitly.
>>>>>>>
>>>>>>> Hi Tao
>>>>>>>
>>>>>>> What exactly does it get wrong?
>>>>>>>
>>>>>>> Thanks
>>>>>>> Andrew
>>>>>>
>>>>>> Hi Andrew,
>>>>>>
>>>>>> BCM54616S is set to RGMII-Fiber (1000Base-X) mode on my platform, and none of the features (1000BaseT/100BaseT/10BaseT) can be detected by genphy_read_abilities(), because the PHY only reports 1000BaseX_Full|Half ability in this mode.
>>>>>>
>>>>> Are you going to use the PHY in copper or fibre mode?
>>>>> In case you use fibre mode, why do you need the copper modes set as supported?
>>>>> Or does the PHY just start in fibre mode and you want to switch it to copper mode?
>>>>
>>>> Hi Heiner,
>>>>
>>>> The phy starts in fiber mode and that's the mode I want.
>>>> My observation is: phydev->link is always 0 (Link status bit is never set in MII_BMSR) by using dynamic ability detection on my machine. I checked phydev->supported and it's set to "AutoNeg | TP | MII | Pause | Asym_Pause" by dynamic ability detection. Is it normal/expected? Or maybe the fix should go to different places? Thank you for your help.
>>>>
>>>
>>> Not sure whether you stated already which kernel version you're using.
>>> There's a brand-new extension to auto-detect 1000BaseX:
>>> f30e33bcdab9 ("net: phy: Add more 1000BaseX support detection")
>>> It's included in the 5.3-rc series.
>>
>> I'm running kernel 5.2.0. Thank you for the sharing and I didn't know the patch. Let me check it out.
>
> I applied above patch and ca72efb6bdc7 ("net: phy: Add detection of 1000BaseX link mode support") to my 5.2.0 tree but got following warning when booting up my machine:
>
> "PHY advertising (0,00000200,000062c0) more modes than genphy supports, some modes not advertised".
>
It's genphy_config_advert complaining which is called from genphy_config_aneg.
genphy_config_aneg deals with the standard Base-T modes. Therefore in your case
most likely you want to provide an own config_aneg callback (in case autoneg
is applicable at all).
> The BCM54616S PHY on my machine only reports 1000-X features in RGMII->1000Base-KX mode. Is it a known problem?
>
> Anyways let me see if I missed some dependency/follow-up patches..
>
>
> Cheers,
>
> Tao
>
Heiner
^ permalink raw reply
* Re: BPF: ETLS: RECV FLOW
From: Shridhar Venkatraman @ 2019-08-04 9:39 UTC (permalink / raw)
To: netdev
In-Reply-To: <CADJe1ZsN8+1brBNdN2VNMp4PRdeYjCC=qaMZALQxOTvPmgJQhA@mail.gmail.com>
Hi,
The eTLS work has BPF integration which is great.
However there is one spot where access to the clear text is not available.
From kernel 4.20 - receiver BPF support added for KTLS.
a. receiver BPF is applied on encrypted message
b. after applying BPF, message is decrypted
c. BPF run logic on the decrypted plain message - can we add this support ?
d. then copy the decrypted message back to userspace.
code flow reference: tls receive message call flow:
--------------------------------------------------------------
tls_sw_recvmsg
__tcp_bpf_recvmsg [ bpf exec function called on encrypted message ]
decrypt_skb_update
decrypt_internal
BPF_PROG_RUN on decrypted plain message - can we add this support ?
skb_copy_datagram_msg [ decrypted message copied back to userspace ]
Thanks
ps: I sent this to the bpf list as I don't know which one it should go to
^ permalink raw reply
* Re: [PATCH iproute2-next] rdma: Add driver QP type string
From: Leon Romanovsky @ 2019-08-04 9:52 UTC (permalink / raw)
To: Gal Pressman; +Cc: Stephen Hemminger, netdev, linux-rdma
In-Reply-To: <20190804080756.58364-1-galpress@amazon.com>
On Sun, Aug 04, 2019 at 11:07:56AM +0300, Gal Pressman wrote:
> RDMA resource tracker now tracks driver QPs as well, add driver QP type
> string to qp_types_to_str function.
>
> Signed-off-by: Gal Pressman <galpress@amazon.com>
> ---
> rdma/res.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
Thanks,
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
^ permalink raw reply
* [PATCH rdma-next v1 0/3] ODP support for mlx5 DC QPs
From: Leon Romanovsky @ 2019-08-04 10:00 UTC (permalink / raw)
To: Doug Ledford, Jason Gunthorpe
Cc: Leon Romanovsky, RDMA mailing list, Michael Guralnik, Moni Shoua,
Saeed Mahameed, linux-netdev
From: Leon Romanovsky <leonro@mellanox.com>
Changelog
v1:
* Fixed alignment to u64 in mlx5-abi.h (Gal P.)
v0:
* https://lore.kernel.org/linux-rdma/20190801122139.25224-1-leon@kernel.org
---------------------------------------------------------------------------------
From Michael,
The series adds support for on-demand paging for DC transport.
Adding handling of DC WQE parsing upon page faults and exposing
capabilities.
As DC is mlx-only transport, the capabilities are exposed to the user
using the direct-verbs mechanism. Namely through the
mlx5dv_query_device.
Thanks
Thanks
Michael Guralnik (3):
IB/mlx5: Query ODP capabilities for DC
IB/mlx5: Expose ODP for DC capabilities to user
IB/mlx5: Add page fault handler for DC initiator WQE
drivers/infiniband/hw/mlx5/main.c | 6 +++++
drivers/infiniband/hw/mlx5/mlx5_ib.h | 1 +
drivers/infiniband/hw/mlx5/odp.c | 27 ++++++++++++++++++-
.../net/ethernet/mellanox/mlx5/core/main.c | 6 +++++
include/linux/mlx5/mlx5_ifc.h | 4 ++-
include/uapi/rdma/mlx5-abi.h | 3 +++
6 files changed, 45 insertions(+), 2 deletions(-)
--
2.20.1
^ permalink raw reply
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