* [PATCH net-next] net: metrics: add proper netlink validation
From: Eric Dumazet @ 2018-06-04 23:46 UTC (permalink / raw)
To: David S . Miller; +Cc: netdev, Eric Dumazet, Eric Dumazet, David Ahern
Before using nla_get_u32(), better make sure the attribute
is of the proper size.
Code recently was changed, but bug has been there from beginning
of git.
BUG: KMSAN: uninit-value in rtnetlink_put_metrics+0x553/0x960 net/core/rtnetlink.c:746
CPU: 1 PID: 14139 Comm: syz-executor6 Not tainted 4.17.0-rc5+ #103
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0x185/0x1d0 lib/dump_stack.c:113
kmsan_report+0x149/0x260 mm/kmsan/kmsan.c:1084
__msan_warning_32+0x6e/0xc0 mm/kmsan/kmsan_instr.c:686
rtnetlink_put_metrics+0x553/0x960 net/core/rtnetlink.c:746
fib_dump_info+0xc42/0x2190 net/ipv4/fib_semantics.c:1361
rtmsg_fib+0x65f/0x8c0 net/ipv4/fib_semantics.c:419
fib_table_insert+0x2314/0x2b50 net/ipv4/fib_trie.c:1287
inet_rtm_newroute+0x210/0x340 net/ipv4/fib_frontend.c:779
rtnetlink_rcv_msg+0xa32/0x1560 net/core/rtnetlink.c:4646
netlink_rcv_skb+0x378/0x600 net/netlink/af_netlink.c:2448
rtnetlink_rcv+0x50/0x60 net/core/rtnetlink.c:4664
netlink_unicast_kernel net/netlink/af_netlink.c:1310 [inline]
netlink_unicast+0x1678/0x1750 net/netlink/af_netlink.c:1336
netlink_sendmsg+0x104f/0x1350 net/netlink/af_netlink.c:1901
sock_sendmsg_nosec net/socket.c:629 [inline]
sock_sendmsg net/socket.c:639 [inline]
___sys_sendmsg+0xec0/0x1310 net/socket.c:2117
__sys_sendmsg net/socket.c:2155 [inline]
__do_sys_sendmsg net/socket.c:2164 [inline]
__se_sys_sendmsg net/socket.c:2162 [inline]
__x64_sys_sendmsg+0x331/0x460 net/socket.c:2162
do_syscall_64+0x152/0x230 arch/x86/entry/common.c:287
entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x455a09
RSP: 002b:00007faae5fd8c68 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007faae5fd96d4 RCX: 0000000000455a09
RDX: 0000000000000000 RSI: 0000000020000000 RDI: 0000000000000013
RBP: 000000000072bea0 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 00000000ffffffff
R13: 00000000000005d0 R14: 00000000006fdc20 R15: 0000000000000000
Uninit was stored to memory at:
kmsan_save_stack_with_flags mm/kmsan/kmsan.c:279 [inline]
kmsan_save_stack mm/kmsan/kmsan.c:294 [inline]
kmsan_internal_chain_origin+0x12b/0x210 mm/kmsan/kmsan.c:685
__msan_chain_origin+0x69/0xc0 mm/kmsan/kmsan_instr.c:529
fib_convert_metrics net/ipv4/fib_semantics.c:1056 [inline]
fib_create_info+0x2d46/0x9dc0 net/ipv4/fib_semantics.c:1150
fib_table_insert+0x3e4/0x2b50 net/ipv4/fib_trie.c:1146
inet_rtm_newroute+0x210/0x340 net/ipv4/fib_frontend.c:779
rtnetlink_rcv_msg+0xa32/0x1560 net/core/rtnetlink.c:4646
netlink_rcv_skb+0x378/0x600 net/netlink/af_netlink.c:2448
rtnetlink_rcv+0x50/0x60 net/core/rtnetlink.c:4664
netlink_unicast_kernel net/netlink/af_netlink.c:1310 [inline]
netlink_unicast+0x1678/0x1750 net/netlink/af_netlink.c:1336
netlink_sendmsg+0x104f/0x1350 net/netlink/af_netlink.c:1901
sock_sendmsg_nosec net/socket.c:629 [inline]
sock_sendmsg net/socket.c:639 [inline]
___sys_sendmsg+0xec0/0x1310 net/socket.c:2117
__sys_sendmsg net/socket.c:2155 [inline]
__do_sys_sendmsg net/socket.c:2164 [inline]
__se_sys_sendmsg net/socket.c:2162 [inline]
__x64_sys_sendmsg+0x331/0x460 net/socket.c:2162
do_syscall_64+0x152/0x230 arch/x86/entry/common.c:287
entry_SYSCALL_64_after_hwframe+0x44/0xa9
Uninit was created at:
kmsan_save_stack_with_flags mm/kmsan/kmsan.c:279 [inline]
kmsan_internal_poison_shadow+0xb8/0x1b0 mm/kmsan/kmsan.c:189
kmsan_kmalloc+0x94/0x100 mm/kmsan/kmsan.c:315
kmsan_slab_alloc+0x10/0x20 mm/kmsan/kmsan.c:322
slab_post_alloc_hook mm/slab.h:446 [inline]
slab_alloc_node mm/slub.c:2753 [inline]
__kmalloc_node_track_caller+0xb32/0x11b0 mm/slub.c:4395
__kmalloc_reserve net/core/skbuff.c:138 [inline]
__alloc_skb+0x2cb/0x9e0 net/core/skbuff.c:206
alloc_skb include/linux/skbuff.h:988 [inline]
netlink_alloc_large_skb net/netlink/af_netlink.c:1182 [inline]
netlink_sendmsg+0x76e/0x1350 net/netlink/af_netlink.c:1876
sock_sendmsg_nosec net/socket.c:629 [inline]
sock_sendmsg net/socket.c:639 [inline]
___sys_sendmsg+0xec0/0x1310 net/socket.c:2117
__sys_sendmsg net/socket.c:2155 [inline]
__do_sys_sendmsg net/socket.c:2164 [inline]
__se_sys_sendmsg net/socket.c:2162 [inline]
__x64_sys_sendmsg+0x331/0x460 net/socket.c:2162
do_syscall_64+0x152/0x230 arch/x86/entry/common.c:287
entry_SYSCALL_64_after_hwframe+0x44/0xa9
Fixes: a919525ad832 ("net: Move fib_convert_metrics to metrics file")
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Cc: David Ahern <dsahern@gmail.com>
---
net/ipv4/fib_semantics.c | 2 ++
net/ipv4/metrics.c | 2 ++
2 files changed, 4 insertions(+)
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index 6608db23f54b6afdac0455650b47d64b1b22b255..9a890be8a0265edb78da225a82e2cac120f2150f 100644
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -717,6 +717,8 @@ bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi)
nla_strlcpy(tmp, nla, sizeof(tmp));
val = tcp_ca_get_key_by_name(fi->fib_net, tmp, &ecn_ca);
} else {
+ if (nla_len(nla) != sizeof(u32)
+ return false;
val = nla_get_u32(nla);
}
diff --git a/net/ipv4/metrics.c b/net/ipv4/metrics.c
index 5121c6475e6b0e9a9a158d4cee473f52cd4d8efe..04311f7067e2e9e3dafb89aa4f8e30dab0fde854 100644
--- a/net/ipv4/metrics.c
+++ b/net/ipv4/metrics.c
@@ -32,6 +32,8 @@ int ip_metrics_convert(struct net *net, struct nlattr *fc_mx, int fc_mx_len,
if (val == TCP_CA_UNSPEC)
return -EINVAL;
} else {
+ if (nla_len(nla) != sizeof(u32))
+ return -EINVAL;
val = nla_get_u32(nla);
}
if (type == RTAX_ADVMSS && val > 65535 - 40)
--
2.17.1.1185.g55be947832-goog
^ permalink raw reply related
* Re: [PATCH bpf-next v2 2/2] samples/bpf: Add xdp_sample_pkts example
From: Daniel Borkmann @ 2018-06-04 23:43 UTC (permalink / raw)
To: Jakub Kicinski, Toke Høiland-Jørgensen; +Cc: netdev
In-Reply-To: <20180604153205.72678576@cakuba.netronome.com>
On 06/05/2018 12:32 AM, Jakub Kicinski wrote:
> On Mon, 04 Jun 2018 18:33:56 +0200, Toke Høiland-Jørgensen wrote:
>> + if (load_bpf_file(filename)) {
>
> Would you mind using libbpf instead of bpf_load.o? I converted some
> samples in be5bca44aa6b ("samples: bpf: convert some XDP samples from
> bpf_load to libbpf"), it's pretty straight forward. Maybe we can kill
> bpf_load.o one day :)
Agreed, we should only be using libbpf going forward.
^ permalink raw reply
* Re: [PATCH bpf-next v2 1/2] trace_helpers.c: Add helpers to poll multiple perf FDs for events
From: Daniel Borkmann @ 2018-06-04 23:42 UTC (permalink / raw)
To: Jakub Kicinski, Toke Høiland-Jørgensen
Cc: netdev, Alexei Starovoitov
In-Reply-To: <20180604152324.6e1115a2@cakuba.netronome.com>
On 06/05/2018 12:26 AM, Jakub Kicinski wrote:
> On Mon, 04 Jun 2018 18:33:56 +0200, Toke Høiland-Jørgensen wrote:
>> This adds two new helper functions to trace_helpers that supports polling
>> multiple perf file descriptors for events. These are used to the XDP
>> perf_event_output example, which needs to work with one perf fd per CPU.
>>
>> Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
>
> Did you take a look at tools/bpf/bpftool/map_perf_ring.c ?
>
> I think the ability to poll multiple FDs could be generally useful and
> therefore better add it to libbpf.c than
> tools/testing/selftests/bpf/trace_helpers.c? I'm not 100% sure myself...
I think for it to land in libbpf this code needs to be more generalized
as it is right now and allowing for more flexibility like pinning RB
processing threads to CPUs, poll handling, etc.
^ permalink raw reply
* Re: [PATCH net-next v2 1/5] net: aquantia: Ethtool based ring size configuration
From: Igor Russkikh @ 2018-06-04 23:35 UTC (permalink / raw)
To: Jakub Kicinski
Cc: David S . Miller, netdev, David Arcari, Pavel Belous,
Anton Mikaev
In-Reply-To: <20180604160024.7a8c7296@cakuba.netronome.com>
>> + mutex_lock(&self->aq_mutex);
>> +
>> if (aq_utils_obj_test(&self->flags, AQ_NIC_FLAGS_IS_NOT_READY))
>> goto err_exit;
>>
>> @@ -175,6 +177,7 @@ static void aq_nic_service_timer_cb(struct timer_list *t)
>> ctimer = max(ctimer / 2, 1);
>>
>> err_exit:
>> + mutex_unlock(&self->aq_mutex);
>> mod_timer(&self->service_timer, jiffies + ctimer);
>> }
>>
>
> This looks like a timer callback from the prototype, I don't think you
> can take mutexes in timer callbacks.
True as well. Eventually, think we may just get rid of mutex inside of this callback.
Mutex then will only serve to prevent possible parallel `ethtool -G` collisions from happening.
BR, Igor
^ permalink raw reply
* AF_XDP. Was: [net-next 00/12][pull request] Intel Wired LAN Driver Updates 2018-06-04
From: Alexei Starovoitov @ 2018-06-04 23:32 UTC (permalink / raw)
To: Alexander Duyck
Cc: David Miller, Björn Töpel, Karlsson, Magnus,
Alexei Starovoitov, Daniel Borkmann, Or Gerlitz, Jeff Kirsher,
Netdev
In-Reply-To: <CAKgT0UdekKLETPDg0Qy58bckFZTST1vXUkOSCz40CoZ1sC-=KA@mail.gmail.com>
On Mon, Jun 04, 2018 at 03:02:31PM -0700, Alexander Duyck wrote:
> On Mon, Jun 4, 2018 at 2:27 PM, David Miller <davem@davemloft.net> wrote:
> > From: Or Gerlitz <gerlitz.or@gmail.com>
> > Date: Tue, 5 Jun 2018 00:11:35 +0300
> >
> >> Just to make sure, is the AF_XDP ZC (Zero Copy) UAPI going to be
> >> merged for this window -- AFAIU from [1], it's still under
> >> examination/development/research for non Intel HWs, am I correct or
> >> this is going to get in now?
> >
> > All of the pending AF_XDP changes will be merged this merge window.
> >
> > I think Intel folks need to review things as fast as possible because
> > I pretty much refuse to revert the series or disable it in Kconfig at
> > this point.
> >
> > Thank you.
>
> My understanding of things is that the current AF_XDP patches were
> going to be updated to have more of a model agnostic API such that
> they would work for either the "typewriter" mode or the descriptor
> ring based approach. The current plan was to have the zero copy
> patches be a follow-on after the vendor agnostic API bits in the
> descriptors and such had been sorted out. I believe you guys have the
> descriptor fixes already right?
>
> In my opinion the i40e code isn't mature enough yet to really go into
> anything other than maybe net-next in a couple weeks. We are going to
> need a while to get adequate testing in order to flush out all the
> bugs and performance regressions we are likely to see coming out of
> this change.
I think the work everyone did in this release cycle increased my confidence
that the way descriptors are defined and the rest of uapi are stable enough
and i40e zero copy bits can land in the next release without uapi changes.
In that sense even if we merge i40e parts now, the other nic vendors
will be in the same situation and may find things that they would like
to improve in uapi.
So I propose we merge the first 7 patches of the last series now and
let 3 remaining i40e patches go via intel trees for the next release.
In the mean time other NIC vendors should start actively working
on AF_XDP support as well.
If somehow uapi would need tweaks, we can still do minor adjustments
since 4.18 won't be released for ~10 weeks.
^ permalink raw reply
* Re: [bpf-next PATCH] bpf: sockmap, fix crash when ipv6 sock is added
From: Daniel Borkmann @ 2018-06-04 23:20 UTC (permalink / raw)
To: John Fastabend, edumazet, ast; +Cc: netdev
In-Reply-To: <60fbb84e-e334-9f53-ced4-170bc2dd21df@gmail.com>
On 06/05/2018 01:08 AM, John Fastabend wrote:
> On 06/04/2018 12:59 PM, Daniel Borkmann wrote:
>> On 06/04/2018 05:21 PM, John Fastabend wrote:
>>> This fixes a crash where we assign tcp_prot to IPv6 sockets instead
>>> of tcpv6_prot.
>>>
>>> Previously we overwrote the sk->prot field with tcp_prot even in the
>>> AF_INET6 case. This patch ensures the correct tcp_prot and tcpv6_prot
>>> are used. Further, only allow ESTABLISHED connections to join the
>>> map per note in TLS ULP,
>>>
>>> /* The TLS ulp is currently supported only for TCP sockets
>>> * in ESTABLISHED state.
>>> * Supporting sockets in LISTEN state will require us
>>> * to modify the accept implementation to clone rather then
>>> * share the ulp context.
>>> */
>>>
>>> Also tested with 'netserver -6' and 'netperf -H [IPv6]' as well as
>>> 'netperf -H [IPv4]'. The ESTABLISHED check resolves the previously
>>> crashing case here.
>>>
>>> Fixes: 174a79ff9515 ("bpf: sockmap with sk redirect support")
>>> Reported-by: syzbot+5c063698bdbfac19f363@syzkaller.appspotmail.com
>>> Signed-off-by: John Fastabend <john.fastabend@gmail.com>
>>> Signed-off-by: Wei Wang <weiwan@google.com>
>>
>> Applied to bpf-next, thanks everyone!
>
> Thanks Daniel, this has the unfortunate side-effect though of
> making it hard to add sockets transitioning from LISTEN into
> ESTABLISHED states to a sockmap. Before this patch we could add
> sockets from the sock_ops event BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB
> to a sock{map|hash}. However, this is before a socket is in established
> state so risked crashing and wasn't valid at all per this thread. So
> I believe its correct to block this action, seeing it will crash a
> system in many (most!) cases.
>
> That said we still would like to support pushing sockets into a
> sock{map|hash} in this case. I thought about adding a new hook but
> we already have a few sock op hooks in the TCP stack so its too bad we
> don't have one that fires after the ESTABLISHED state has transitioned.
> Right now I'm looking into seeing if the following would have any
> issues,
>
> --- a/net/ipv4/tcp.c
> +++ b/net/ipv4/tcp.c
> @@ -2206,9 +2206,6 @@ void tcp_set_state(struct sock *sk, int state)
> BUILD_BUG_ON((int)BPF_TCP_NEW_SYN_RECV != (int)TCP_NEW_SYN_RECV);
> BUILD_BUG_ON((int)BPF_TCP_MAX_STATES != (int)TCP_MAX_STATES);
>
> - if (BPF_SOCK_OPS_TEST_FLAG(tcp_sk(sk), BPF_SOCK_OPS_STATE_CB_FLAG))
> - tcp_call_bpf_2arg(sk, BPF_SOCK_OPS_STATE_CB, oldstate, state);
> -
> switch (state) {
> case TCP_ESTABLISHED:
> if (oldstate != TCP_ESTABLISHED)
> @@ -2234,6 +2231,9 @@ void tcp_set_state(struct sock *sk, int state)
> */
> inet_sk_state_store(sk, state);
>
> + if (BPF_SOCK_OPS_TEST_FLAG(tcp_sk(sk), BPF_SOCK_OPS_STATE_CB_FLAG))
> + tcp_call_bpf_2arg(sk, BPF_SOCK_OPS_STATE_CB, oldstate, state);
> +
> #ifdef STATE_TRACE
> SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n", sk, statename[oldstate], statename[state]);
> #endif
>
> This would change the call hook slightly, moving it to after the state
> change. However unless the unhash is some how visible from the bpf program
> I don't think it should impact existing BPF programs.
Hmm, the current fix also breaks compilation when IPv6 is compiled out, so I had
to take it out for now. :-( I think this needs similar workaround as in kTLS case
in tls_init(). Given this and your above seen side-effect, lets respin all with a
clean fix.
tree: https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git master
head: 4e1f687e302835e45e2f296392f21cfeb5671303
commit: 4e1f687e302835e45e2f296392f21cfeb5671303 [3/3] bpf: sockmap, fix crash when ipv6 sock is added
config: i386-randconfig-a0-06041847 (attached as .config)
compiler: gcc-4.9 (Debian 4.9.4-2) 4.9.4
reproduce:
git checkout 4e1f687e302835e45e2f296392f21cfeb5671303
# save the attached .config to linux build tree
make ARCH=i386
All errors (new ones prefixed by >>):
kernel/bpf/sockmap.o: In function `bpf_tcp_ulp_register':
>> kernel/bpf/sockmap.c:1127: undefined reference to `tcpv6_prot'
>> kernel/bpf/sockmap.c:1127: undefined reference to `tcpv6_prot'
vim +1127 kernel/bpf/sockmap.c
1122
1123 static int bpf_tcp_ulp_register(void)
1124 {
1125 tcp_bpf_proto = tcp_prot;
1126 tcp_bpf_proto.close = bpf_tcp_close;
> 1127 tcpv6_bpf_proto = tcpv6_prot;
1128 tcpv6_bpf_proto.close = bpf_tcp_close;
1129 /* Once BPF TX ULP is registered it is never unregistered. It
1130 * will be in the ULP list for the lifetime of the system. Doing
1131 * duplicate registers is not a problem.
1132 */
1133 return tcp_register_ulp(&bpf_tcp_ulp_ops);
1134 }
1135
Thanks,
Daniel
^ permalink raw reply
* Re: [PATCH net] ipmr: fix error path when mr_table_alloc fails
From: kbuild test robot @ 2018-06-05 7:29 UTC (permalink / raw)
To: Sabrina Dubroca
Cc: kbuild-all, netdev, Sabrina Dubroca, Eric Dumazet,
Nikolay Aleksandrov, Yuval Mintz, Ivan Vecera
In-Reply-To: <ffc3366697c8d8789cbe314b6944b910eceb38e7.1528112758.git.sd@queasysnail.net>
[-- Attachment #1: Type: text/plain, Size: 3404 bytes --]
Hi Sabrina,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on net/master]
url: https://github.com/0day-ci/linux/commits/Sabrina-Dubroca/ipmr-fix-error-path-when-mr_table_alloc-fails/20180605-060837
config: x86_64-randconfig-x006-201822 (attached as .config)
compiler: gcc-7 (Debian 7.3.0-16) 7.3.0
reproduce:
# save the attached .config to linux build tree
make ARCH=x86_64
Note: it may well be a FALSE warning. FWIW you are at least aware of it now.
http://gcc.gnu.org/wiki/Better_Uninitialized_Warnings
All warnings (new ones prefixed by >>):
In file included from arch/x86/include/asm/current.h:5:0,
from include/linux/sched.h:12,
from include/linux/uaccess.h:5,
from net/ipv6/ip6mr.c:19:
net/ipv6/ip6mr.c: In function 'ip6_mroute_setsockopt':
>> include/linux/compiler.h:177:26: warning: 'mrt' may be used uninitialized in this function [-Wmaybe-uninitialized]
case 8: *(__u64 *)res = *(volatile __u64 *)p; break; \
^
net/ipv6/ip6mr.c:1751:20: note: 'mrt' was declared here
struct mr_table *mrt;
^~~
--
In file included from arch/x86/include/asm/current.h:5:0,
from include/linux/sched.h:12,
from include/linux/uaccess.h:5,
from net//ipv6/ip6mr.c:19:
net//ipv6/ip6mr.c: In function 'ip6_mroute_setsockopt':
>> include/linux/compiler.h:177:26: warning: 'mrt' may be used uninitialized in this function [-Wmaybe-uninitialized]
case 8: *(__u64 *)res = *(volatile __u64 *)p; break; \
^
net//ipv6/ip6mr.c:1751:20: note: 'mrt' was declared here
struct mr_table *mrt;
^~~
vim +/mrt +177 include/linux/compiler.h
230fa253 Christian Borntraeger 2014-11-25 170
d976441f Andrey Ryabinin 2015-10-19 171 #define __READ_ONCE_SIZE \
d976441f Andrey Ryabinin 2015-10-19 172 ({ \
d976441f Andrey Ryabinin 2015-10-19 173 switch (size) { \
d976441f Andrey Ryabinin 2015-10-19 174 case 1: *(__u8 *)res = *(volatile __u8 *)p; break; \
d976441f Andrey Ryabinin 2015-10-19 175 case 2: *(__u16 *)res = *(volatile __u16 *)p; break; \
d976441f Andrey Ryabinin 2015-10-19 176 case 4: *(__u32 *)res = *(volatile __u32 *)p; break; \
d976441f Andrey Ryabinin 2015-10-19 @177 case 8: *(__u64 *)res = *(volatile __u64 *)p; break; \
d976441f Andrey Ryabinin 2015-10-19 178 default: \
d976441f Andrey Ryabinin 2015-10-19 179 barrier(); \
d976441f Andrey Ryabinin 2015-10-19 180 __builtin_memcpy((void *)res, (const void *)p, size); \
d976441f Andrey Ryabinin 2015-10-19 181 barrier(); \
d976441f Andrey Ryabinin 2015-10-19 182 } \
d976441f Andrey Ryabinin 2015-10-19 183 })
d976441f Andrey Ryabinin 2015-10-19 184
:::::: The code at line 177 was first introduced by commit
:::::: d976441f44bc5d48635d081d277aa76556ffbf8b compiler, atomics, kasan: Provide READ_ONCE_NOCHECK()
:::::: TO: Andrey Ryabinin <aryabinin@virtuozzo.com>
:::::: CC: Ingo Molnar <mingo@kernel.org>
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 29370 bytes --]
^ permalink raw reply
* Re: [bpf-next PATCH] bpf: sockmap, fix crash when ipv6 sock is added
From: John Fastabend @ 2018-06-04 23:08 UTC (permalink / raw)
To: Daniel Borkmann, edumazet, ast; +Cc: netdev
In-Reply-To: <ce73fcc1-15c4-7384-10dd-aeb5fbc159ae@iogearbox.net>
On 06/04/2018 12:59 PM, Daniel Borkmann wrote:
> On 06/04/2018 05:21 PM, John Fastabend wrote:
>> This fixes a crash where we assign tcp_prot to IPv6 sockets instead
>> of tcpv6_prot.
>>
>> Previously we overwrote the sk->prot field with tcp_prot even in the
>> AF_INET6 case. This patch ensures the correct tcp_prot and tcpv6_prot
>> are used. Further, only allow ESTABLISHED connections to join the
>> map per note in TLS ULP,
>>
>> /* The TLS ulp is currently supported only for TCP sockets
>> * in ESTABLISHED state.
>> * Supporting sockets in LISTEN state will require us
>> * to modify the accept implementation to clone rather then
>> * share the ulp context.
>> */
>>
>> Also tested with 'netserver -6' and 'netperf -H [IPv6]' as well as
>> 'netperf -H [IPv4]'. The ESTABLISHED check resolves the previously
>> crashing case here.
>>
>> Fixes: 174a79ff9515 ("bpf: sockmap with sk redirect support")
>> Reported-by: syzbot+5c063698bdbfac19f363@syzkaller.appspotmail.com
>> Signed-off-by: John Fastabend <john.fastabend@gmail.com>
>> Signed-off-by: Wei Wang <weiwan@google.com>
>
> Applied to bpf-next, thanks everyone!
>
Thanks Daniel, this has the unfortunate side-effect though of
making it hard to add sockets transitioning from LISTEN into
ESTABLISHED states to a sockmap. Before this patch we could add
sockets from the sock_ops event BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB
to a sock{map|hash}. However, this is before a socket is in established
state so risked crashing and wasn't valid at all per this thread. So
I believe its correct to block this action, seeing it will crash a
system in many (most!) cases.
That said we still would like to support pushing sockets into a
sock{map|hash} in this case. I thought about adding a new hook but
we already have a few sock op hooks in the TCP stack so its too bad we
don't have one that fires after the ESTABLISHED state has transitioned.
Right now I'm looking into seeing if the following would have any
issues,
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -2206,9 +2206,6 @@ void tcp_set_state(struct sock *sk, int state)
BUILD_BUG_ON((int)BPF_TCP_NEW_SYN_RECV != (int)TCP_NEW_SYN_RECV);
BUILD_BUG_ON((int)BPF_TCP_MAX_STATES != (int)TCP_MAX_STATES);
- if (BPF_SOCK_OPS_TEST_FLAG(tcp_sk(sk), BPF_SOCK_OPS_STATE_CB_FLAG))
- tcp_call_bpf_2arg(sk, BPF_SOCK_OPS_STATE_CB, oldstate, state);
-
switch (state) {
case TCP_ESTABLISHED:
if (oldstate != TCP_ESTABLISHED)
@@ -2234,6 +2231,9 @@ void tcp_set_state(struct sock *sk, int state)
*/
inet_sk_state_store(sk, state);
+ if (BPF_SOCK_OPS_TEST_FLAG(tcp_sk(sk), BPF_SOCK_OPS_STATE_CB_FLAG))
+ tcp_call_bpf_2arg(sk, BPF_SOCK_OPS_STATE_CB, oldstate, state);
+
#ifdef STATE_TRACE
SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n", sk, statename[oldstate], statename[state]);
#endif
This would change the call hook slightly, moving it to after the state
change. However unless the unhash is some how visible from the bpf program
I don't think it should impact existing BPF programs.
Thanks,
John
^ permalink raw reply
* Re: [PATCH net-next v11 04/10] netdev: cavium: octeon: Add Octeon III BGX Ports
From: Andrew Lunn @ 2018-06-04 23:07 UTC (permalink / raw)
To: Steven J. Hill; +Cc: netdev, Carlos Munoz
In-Reply-To: <1528149617-8964-5-git-send-email-steven.hill@cavium.com>
> +static int bgx_port_get_qlm_speed(struct bgx_port_priv *priv, int qlm)
> +{
> + enum lane_mode lmode;
> + u64 data;
> +
> + data = oct_csr_read(GSER_LANE_MODE(priv->node, qlm));
> + lmode = data & 0xf;
> +
> + switch (lmode) {
> + case R_25G_REFCLK100:
> + return 2500;
> + case R_5G_REFCLK100:
> + return 5000;
> + case R_8G_REFCLK100:
> + return 8000;
> + case R_125G_REFCLK15625_KX:
> + return 1250;
> + case R_3125G_REFCLK15625_XAUI:
> + return 3125;
> + case R_103125G_REFCLK15625_KR:
> + return 10312;
> + case R_125G_REFCLK15625_SGMII:
> + return 1250;
> + case R_5G_REFCLK15625_QSGMII:
> + return 5000;
> + case R_625G_REFCLK15625_RXAUI:
> + return 6250;
> + case R_25G_REFCLK125:
> + return 2500;
> + case R_5G_REFCLK125:
> + return 5000;
> + case R_8G_REFCLK125:
> + return 8000;
> + default:
> + return 0;
> + }
> + struct port_status status;
> + int speed;
> +
> + /* The simulator always uses a 1Gbps full duplex port */
> + if (octeon_is_simulation()) {
> + status.link = 1;
> + status.duplex = DUPLEX_FULL;
> + status.speed = 1000;
> + } else {
> + /* Use the qlm speed */
> + speed = bgx_port_get_qlm_speed(priv, priv->qlm);
> + status.link = 1;
> + status.duplex = DUPLEX_FULL;
> + status.speed = speed * 8 / 10;
> + }
Hi Steve
That looks like it gives some odd speeds.
2500 * 8 / 10 = 2Gbps
5000 * 8 / 10 = 4Gbps
8000 * 8 / 10 = 6.4Gbps
10312 * 8 /10 = 8.249Gbps
Is this correct. We are more used to 10/100/1000/2.5G/5G/10G/40G/100G.
Andrew
^ permalink raw reply
* Re: [PATCH net-next v2 1/5] net: aquantia: Ethtool based ring size configuration
From: Jakub Kicinski @ 2018-06-04 23:00 UTC (permalink / raw)
To: Igor Russkikh
Cc: David S . Miller, netdev, David Arcari, Pavel Belous,
Anton Mikaev
In-Reply-To: <5f24d3f28fe0014c33fde9a20af547cd3b9f8d9d.1528150073.git.igor.russkikh@aquantia.com>
On Tue, 5 Jun 2018 01:30:15 +0300, Igor Russkikh wrote:
> @@ -158,6 +158,8 @@ static void aq_nic_service_timer_cb(struct timer_list *t)
> int ctimer = AQ_CFG_SERVICE_TIMER_INTERVAL;
> int err = 0;
>
> + mutex_lock(&self->aq_mutex);
> +
> if (aq_utils_obj_test(&self->flags, AQ_NIC_FLAGS_IS_NOT_READY))
> goto err_exit;
>
> @@ -175,6 +177,7 @@ static void aq_nic_service_timer_cb(struct timer_list *t)
> ctimer = max(ctimer / 2, 1);
>
> err_exit:
> + mutex_unlock(&self->aq_mutex);
> mod_timer(&self->service_timer, jiffies + ctimer);
> }
>
This looks like a timer callback from the prototype, I don't think you
can take mutexes in timer callbacks.
^ permalink raw reply
* Re: [PATCH net-next 2/3] netdevsim: Add extack error message for devlink reload
From: Jakub Kicinski @ 2018-06-04 22:47 UTC (permalink / raw)
To: dsahern; +Cc: netdev, idosch, jiri, David Ahern
In-Reply-To: <20180604221503.20329-3-dsahern@kernel.org>
On Mon, 4 Jun 2018 15:15:02 -0700, dsahern@kernel.org wrote:
> From: David Ahern <dsahern@gmail.com>
>
> devlink reset command can fail if a FIB resource limit is set to a value
> lower than the current occupancy. Return a proper message indicating the
> reason for the failure.
>
> $ devlink resource sh netdevsim/netdevsim0
> netdevsim/netdevsim0:
> name IPv4 size unlimited unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none
> resources:
> name fib size unlimited occ 43 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none
> name fib-rules size unlimited occ 4 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none
> name IPv6 size unlimited unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none
> resources:
> name fib size unlimited occ 54 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none
> name fib-rules size unlimited occ 3 unit entry size_min 0 size_max unlimited size_gran 1 dpipe_tables none
>
> $ devlink resource set netdevsim/netdevsim0 path /IPv4/fib size 40
>
> $ devlink dev reload netdevsim/netdevsim0
> Error: netdevsim: New size is less than current occupancy.
> devlink answers: Invalid argument
>
> Signed-off-by: David Ahern <dsahern@gmail.com>
Acked-by: Jakub Kicinski <jakub.kicinski@netronome.com>
The entire set looks very useful, thanks!
^ permalink raw reply
* Re: [PATCH bpf-next v3 05/11] bpf: avoid retpoline for lookup/update/delete calls on maps
From: David Ahern @ 2018-06-04 22:36 UTC (permalink / raw)
To: Jakub Kicinski, Phil Sutter
Cc: Jesper Dangaard Brouer, Daniel Borkmann, alexei.starovoitov,
netdev, Jakub Kicinski, Quentin Monnet
In-Reply-To: <20180604112524.1d5732e8@cakuba.netronome.com>
On 6/4/18 11:25 AM, Jakub Kicinski wrote:
> that, and others can use completions. I personally think Quentin did
> an awesome job on the completions, they cover the entire syntax unlike
> the iproute2 ones and we intend to keep them that way!
iproute2 patches for completions would be welcomed if anyone has the time.
^ permalink raw reply
* Re: [PATCH bpf-next v2 2/2] samples/bpf: Add xdp_sample_pkts example
From: Jakub Kicinski @ 2018-06-04 22:32 UTC (permalink / raw)
To: Toke Høiland-Jørgensen; +Cc: netdev
In-Reply-To: <152813003614.3465.11049830599815911223.stgit@alrua-kau>
On Mon, 04 Jun 2018 18:33:56 +0200, Toke Høiland-Jørgensen wrote:
> + if (load_bpf_file(filename)) {
Would you mind using libbpf instead of bpf_load.o? I converted some
samples in be5bca44aa6b ("samples: bpf: convert some XDP samples from
bpf_load to libbpf"), it's pretty straight forward. Maybe we can kill
bpf_load.o one day :)
^ permalink raw reply
* [PATCH net-next v2 5/5] net: aquantia: bump driver version
From: Igor Russkikh @ 2018-06-04 22:30 UTC (permalink / raw)
To: David S . Miller; +Cc: netdev, David Arcari, Pavel Belous, Igor Russkikh
In-Reply-To: <cover.1528150073.git.igor.russkikh@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
---
drivers/net/ethernet/aquantia/atlantic/ver.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/aquantia/atlantic/ver.h b/drivers/net/ethernet/aquantia/atlantic/ver.h
index a445de6..94efc64 100644
--- a/drivers/net/ethernet/aquantia/atlantic/ver.h
+++ b/drivers/net/ethernet/aquantia/atlantic/ver.h
@@ -12,8 +12,8 @@
#define NIC_MAJOR_DRIVER_VERSION 2
#define NIC_MINOR_DRIVER_VERSION 0
-#define NIC_BUILD_DRIVER_VERSION 2
-#define NIC_REVISION_DRIVER_VERSION 1
+#define NIC_BUILD_DRIVER_VERSION 3
+#define NIC_REVISION_DRIVER_VERSION 0
#define AQ_CFG_DRV_VERSION_SUFFIX "-kern"
--
2.7.4
^ permalink raw reply related
* [PATCH net-next v2 4/5] net: aquantia: Add renegotiate ethtool operation support
From: Igor Russkikh @ 2018-06-04 22:30 UTC (permalink / raw)
To: David S . Miller
Cc: netdev, David Arcari, Pavel Belous, Igor Russkikh, Anton Mikaev
In-Reply-To: <cover.1528150073.git.igor.russkikh@aquantia.com>
From: Anton Mikaev <amikaev@aquantia.com>
Adds ethtool -r|--negotiate operation support. It triggers special
control bit on FW interface causing FW to restart link negotiation.
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
Signed-off-by: Anton Mikaev <amikaev@aquantia.com>
---
.../net/ethernet/aquantia/atlantic/aq_ethtool.c | 14 +++++++++
drivers/net/ethernet/aquantia/atlantic/aq_hw.h | 2 ++
.../aquantia/atlantic/hw_atl/hw_atl_utils.h | 35 ++++++++++++++++++++++
.../aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c | 12 ++++++++
4 files changed, 63 insertions(+)
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c b/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
index ec9ed44..7058f130 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
@@ -285,6 +285,19 @@ static int aq_ethtool_set_coalesce(struct net_device *ndev,
return aq_nic_update_interrupt_moderation_settings(aq_nic);
}
+static int aq_ethtool_nway_reset(struct net_device *ndev)
+{
+ struct aq_nic_s *aq_nic = netdev_priv(ndev);
+
+ if (unlikely(!aq_nic->aq_fw_ops->renegotiate))
+ return -EOPNOTSUPP;
+
+ if (netif_running(ndev))
+ return aq_nic->aq_fw_ops->renegotiate(aq_nic->aq_hw);
+
+ return 0;
+}
+
static void aq_ethtool_get_pauseparam(struct net_device *ndev,
struct ethtool_pauseparam *pause)
{
@@ -394,6 +407,7 @@ const struct ethtool_ops aq_ethtool_ops = {
.get_drvinfo = aq_ethtool_get_drvinfo,
.get_strings = aq_ethtool_get_strings,
.get_rxfh_indir_size = aq_ethtool_get_rss_indir_size,
+ .nway_reset = aq_ethtool_nway_reset,
.get_ringparam = aq_get_ringparam,
.set_ringparam = aq_set_ringparam,
.get_pauseparam = aq_ethtool_get_pauseparam,
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_hw.h b/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
index 3aa36d5..1a51152 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
@@ -212,6 +212,8 @@ struct aq_fw_ops {
int (*reset)(struct aq_hw_s *self);
+ int (*renegotiate)(struct aq_hw_s *self);
+
int (*get_mac_permanent)(struct aq_hw_s *self, u8 *mac);
int (*set_link_speed)(struct aq_hw_s *self, u32 speed);
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.h b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.h
index cd8f18f..b875590 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.h
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.h
@@ -239,6 +239,41 @@ enum hw_atl_fw2x_caps_hi {
CAPS_HI_TRANSACTION_ID,
};
+enum hw_atl_fw2x_ctrl {
+ CTRL_RESERVED1 = 0x00,
+ CTRL_RESERVED2,
+ CTRL_RESERVED3,
+ CTRL_PAUSE,
+ CTRL_ASYMMETRIC_PAUSE,
+ CTRL_RESERVED4,
+ CTRL_RESERVED5,
+ CTRL_RESERVED6,
+ CTRL_1GBASET_FD_EEE,
+ CTRL_2P5GBASET_FD_EEE,
+ CTRL_5GBASET_FD_EEE,
+ CTRL_10GBASET_FD_EEE,
+ CTRL_THERMAL_SHUTDOWN,
+ CTRL_PHY_LOGS,
+ CTRL_EEE_AUTO_DISABLE,
+ CTRL_PFC,
+ CTRL_WAKE_ON_LINK,
+ CTRL_CABLE_DIAG,
+ CTRL_TEMPERATURE,
+ CTRL_DOWNSHIFT,
+ CTRL_PTP_AVB,
+ CTRL_RESERVED7,
+ CTRL_LINK_DROP,
+ CTRL_SLEEP_PROXY,
+ CTRL_WOL,
+ CTRL_MAC_STOP,
+ CTRL_EXT_LOOPBACK,
+ CTRL_INT_LOOPBACK,
+ CTRL_RESERVED8,
+ CTRL_WOL_TIMER,
+ CTRL_STATISTICS,
+ CTRL_FORCE_RECONNECT,
+};
+
struct aq_hw_s;
struct aq_fw_ops;
struct aq_hw_caps_s;
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
index d2d030a..1935fd6 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
@@ -215,6 +215,17 @@ static int aq_fw2x_update_stats(struct aq_hw_s *self)
return hw_atl_utils_update_stats(self);
}
+static int aq_fw2x_renegotiate(struct aq_hw_s *self)
+{
+ u32 mpi_opts = aq_hw_read_reg(self, HW_ATL_FW2X_MPI_CONTROL2_ADDR);
+
+ mpi_opts |= BIT(CTRL_FORCE_RECONNECT);
+
+ aq_hw_write_reg(self, HW_ATL_FW2X_MPI_CONTROL2_ADDR, mpi_opts);
+
+ return 0;
+}
+
static int aq_fw2x_set_flow_control(struct aq_hw_s *self)
{
u32 mpi_state = aq_hw_read_reg(self, HW_ATL_FW2X_MPI_CONTROL2_ADDR);
@@ -230,6 +241,7 @@ const struct aq_fw_ops aq_fw_2x_ops = {
.init = aq_fw2x_init,
.deinit = aq_fw2x_deinit,
.reset = NULL,
+ .renegotiate = aq_fw2x_renegotiate,
.get_mac_permanent = aq_fw2x_get_mac_permanent,
.set_link_speed = aq_fw2x_set_link_speed,
.set_state = aq_fw2x_set_state,
--
2.7.4
^ permalink raw reply related
* [PATCH net-next v2 3/5] net: aquantia: Implement rx/tx flow control ethtools callback
From: Igor Russkikh @ 2018-06-04 22:30 UTC (permalink / raw)
To: David S . Miller; +Cc: netdev, David Arcari, Pavel Belous, Igor Russkikh
In-Reply-To: <cover.1528150073.git.igor.russkikh@aquantia.com>
Runtime change of pause frame configuration (rx/tx flow control)
via ethtool.
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
---
.../net/ethernet/aquantia/atlantic/aq_ethtool.c | 42 ++++++++++++++++++++++
drivers/net/ethernet/aquantia/atlantic/aq_nic.c | 6 +++-
.../aquantia/atlantic/hw_atl/hw_atl_utils.c | 1 +
.../aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c | 26 ++++++++++++++
4 files changed, 74 insertions(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c b/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
index 97f42d1..ec9ed44 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
@@ -285,6 +285,46 @@ static int aq_ethtool_set_coalesce(struct net_device *ndev,
return aq_nic_update_interrupt_moderation_settings(aq_nic);
}
+static void aq_ethtool_get_pauseparam(struct net_device *ndev,
+ struct ethtool_pauseparam *pause)
+{
+ struct aq_nic_s *aq_nic = netdev_priv(ndev);
+
+ pause->autoneg = 0;
+
+ if (aq_nic->aq_hw->aq_nic_cfg->flow_control & AQ_NIC_FC_RX)
+ pause->rx_pause = 1;
+ if (aq_nic->aq_hw->aq_nic_cfg->flow_control & AQ_NIC_FC_TX)
+ pause->tx_pause = 1;
+}
+
+static int aq_ethtool_set_pauseparam(struct net_device *ndev,
+ struct ethtool_pauseparam *pause)
+{
+ struct aq_nic_s *aq_nic = netdev_priv(ndev);
+ int err = 0;
+
+ if (!aq_nic->aq_fw_ops->set_flow_control)
+ return -EOPNOTSUPP;
+
+ if (pause->autoneg == AUTONEG_ENABLE)
+ return -EOPNOTSUPP;
+
+ if (pause->rx_pause)
+ aq_nic->aq_hw->aq_nic_cfg->flow_control |= AQ_NIC_FC_RX;
+ else
+ aq_nic->aq_hw->aq_nic_cfg->flow_control &= ~AQ_NIC_FC_RX;
+
+ if (pause->tx_pause)
+ aq_nic->aq_hw->aq_nic_cfg->flow_control |= AQ_NIC_FC_TX;
+ else
+ aq_nic->aq_hw->aq_nic_cfg->flow_control &= ~AQ_NIC_FC_TX;
+
+ err = aq_nic->aq_fw_ops->set_flow_control(aq_nic->aq_hw);
+
+ return err;
+}
+
static void aq_get_ringparam(struct net_device *ndev,
struct ethtool_ringparam *ring)
{
@@ -356,6 +396,8 @@ const struct ethtool_ops aq_ethtool_ops = {
.get_rxfh_indir_size = aq_ethtool_get_rss_indir_size,
.get_ringparam = aq_get_ringparam,
.set_ringparam = aq_set_ringparam,
+ .get_pauseparam = aq_ethtool_get_pauseparam,
+ .set_pauseparam = aq_ethtool_set_pauseparam,
.get_rxfh_key_size = aq_ethtool_get_rss_key_size,
.get_rxfh = aq_ethtool_get_rss,
.get_rxnfc = aq_ethtool_get_rxnfc,
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_nic.c b/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
index 6721ffa..f97e0ba 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
@@ -766,10 +766,14 @@ void aq_nic_get_link_ksettings(struct aq_nic_s *self,
ethtool_link_ksettings_add_link_mode(cmd, advertising,
100baseT_Full);
- if (self->aq_nic_cfg.flow_control)
+ if (self->aq_nic_cfg.flow_control & AQ_NIC_FC_RX)
ethtool_link_ksettings_add_link_mode(cmd, advertising,
Pause);
+ if (self->aq_nic_cfg.flow_control & AQ_NIC_FC_TX)
+ ethtool_link_ksettings_add_link_mode(cmd, advertising,
+ Asym_Pause);
+
if (self->aq_nic_cfg.aq_hw_caps->media_type == AQ_HW_MEDIA_TYPE_FIBRE)
ethtool_link_ksettings_add_link_mode(cmd, advertising, FIBRE);
else
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c
index 9d0a96d..e1feba5 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c
@@ -834,4 +834,5 @@ const struct aq_fw_ops aq_fw_1x_ops = {
.set_state = hw_atl_utils_mpi_set_state,
.update_link_status = hw_atl_utils_mpi_get_link_status,
.update_stats = hw_atl_utils_update_stats,
+ .set_flow_control = NULL,
};
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
index a4ac592..d2d030a 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
@@ -87,6 +87,19 @@ static int aq_fw2x_set_link_speed(struct aq_hw_s *self, u32 speed)
return 0;
}
+static void aq_fw2x_set_mpi_flow_control(struct aq_hw_s *self, u32 *mpi_state)
+{
+ if (self->aq_nic_cfg->flow_control & AQ_NIC_FC_RX)
+ *mpi_state |= BIT(CAPS_HI_PAUSE);
+ else
+ *mpi_state &= ~BIT(CAPS_HI_PAUSE);
+
+ if (self->aq_nic_cfg->flow_control & AQ_NIC_FC_TX)
+ *mpi_state |= BIT(CAPS_HI_ASYMMETRIC_PAUSE);
+ else
+ *mpi_state &= ~BIT(CAPS_HI_ASYMMETRIC_PAUSE);
+}
+
static int aq_fw2x_set_state(struct aq_hw_s *self,
enum hal_atl_utils_fw_state_e state)
{
@@ -95,6 +108,7 @@ static int aq_fw2x_set_state(struct aq_hw_s *self,
switch (state) {
case MPI_INIT:
mpi_state &= ~BIT(CAPS_HI_LINK_DROP);
+ aq_fw2x_set_mpi_flow_control(self, &mpi_state);
break;
case MPI_DEINIT:
mpi_state |= BIT(CAPS_HI_LINK_DROP);
@@ -201,6 +215,17 @@ static int aq_fw2x_update_stats(struct aq_hw_s *self)
return hw_atl_utils_update_stats(self);
}
+static int aq_fw2x_set_flow_control(struct aq_hw_s *self)
+{
+ u32 mpi_state = aq_hw_read_reg(self, HW_ATL_FW2X_MPI_CONTROL2_ADDR);
+
+ aq_fw2x_set_mpi_flow_control(self, &mpi_state);
+
+ aq_hw_write_reg(self, HW_ATL_FW2X_MPI_CONTROL2_ADDR, mpi_state);
+
+ return 0;
+}
+
const struct aq_fw_ops aq_fw_2x_ops = {
.init = aq_fw2x_init,
.deinit = aq_fw2x_deinit,
@@ -210,4 +235,5 @@ const struct aq_fw_ops aq_fw_2x_ops = {
.set_state = aq_fw2x_set_state,
.update_link_status = aq_fw2x_update_link_status,
.update_stats = aq_fw2x_update_stats,
+ .set_flow_control = aq_fw2x_set_flow_control,
};
--
2.7.4
^ permalink raw reply related
* [PATCH net-next v2 2/5] net: aquantia: Improve adapter init/deinit logic
From: Igor Russkikh @ 2018-06-04 22:30 UTC (permalink / raw)
To: David S . Miller; +Cc: netdev, David Arcari, Pavel Belous, Igor Russkikh
In-Reply-To: <cover.1528150073.git.igor.russkikh@aquantia.com>
We now pass link drop status to FW on init/deinit. This is required
to inform FW that driver took/released a control on link.
FW then will manage its own state and device power profile based
on this information. To improve management we remove mpi_set
function which ambiguously took both state and speed parameters.
Deinit callback is now a part of FW ops, as it actually manages the FW.
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
---
drivers/net/ethernet/aquantia/atlantic/aq_hw.h | 9 ++--
drivers/net/ethernet/aquantia/atlantic/aq_nic.c | 2 +-
.../ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c | 1 -
.../ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c | 1 -
.../aquantia/atlantic/hw_atl/hw_atl_utils.c | 53 ++++++++++++----------
.../aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c | 31 ++++++++++++-
6 files changed, 66 insertions(+), 31 deletions(-)
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_hw.h b/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
index 904cdfd..3aa36d5 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
@@ -202,25 +202,28 @@ struct aq_hw_ops {
int (*hw_get_fw_version)(struct aq_hw_s *self, u32 *fw_version);
- int (*hw_deinit)(struct aq_hw_s *self);
-
int (*hw_set_power)(struct aq_hw_s *self, unsigned int power_state);
};
struct aq_fw_ops {
int (*init)(struct aq_hw_s *self);
+ int (*deinit)(struct aq_hw_s *self);
+
int (*reset)(struct aq_hw_s *self);
int (*get_mac_permanent)(struct aq_hw_s *self, u8 *mac);
int (*set_link_speed)(struct aq_hw_s *self, u32 speed);
- int (*set_state)(struct aq_hw_s *self, enum hal_atl_utils_fw_state_e state);
+ int (*set_state)(struct aq_hw_s *self,
+ enum hal_atl_utils_fw_state_e state);
int (*update_link_status)(struct aq_hw_s *self);
int (*update_stats)(struct aq_hw_s *self);
+
+ int (*set_flow_control)(struct aq_hw_s *self);
};
#endif /* AQ_HW_H */
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_nic.c b/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
index a5ccfde..6721ffa 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
@@ -884,7 +884,7 @@ void aq_nic_deinit(struct aq_nic_s *self)
aq_vec_deinit(aq_vec);
if (self->power_state == AQ_HW_POWER_STATE_D0) {
- (void)self->aq_hw_ops->hw_deinit(self->aq_hw);
+ (void)self->aq_fw_ops->deinit(self->aq_hw);
} else {
(void)self->aq_hw_ops->hw_set_power(self->aq_hw,
self->power_state);
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c
index 7fd6a7e..ed7fe6f 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c
@@ -877,7 +877,6 @@ static int hw_atl_a0_hw_ring_rx_stop(struct aq_hw_s *self,
const struct aq_hw_ops hw_atl_ops_a0 = {
.hw_set_mac_address = hw_atl_a0_hw_mac_addr_set,
.hw_init = hw_atl_a0_hw_init,
- .hw_deinit = hw_atl_utils_hw_deinit,
.hw_set_power = hw_atl_utils_hw_set_power,
.hw_reset = hw_atl_a0_hw_reset,
.hw_start = hw_atl_a0_hw_start,
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c
index 4ea15b9..9dd4f49 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c
@@ -935,7 +935,6 @@ static int hw_atl_b0_hw_ring_rx_stop(struct aq_hw_s *self,
const struct aq_hw_ops hw_atl_ops_b0 = {
.hw_set_mac_address = hw_atl_b0_hw_mac_addr_set,
.hw_init = hw_atl_b0_hw_init,
- .hw_deinit = hw_atl_utils_hw_deinit,
.hw_set_power = hw_atl_utils_hw_set_power,
.hw_reset = hw_atl_b0_hw_reset,
.hw_start = hw_atl_b0_hw_start,
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c
index e652d86..9d0a96d 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils.c
@@ -30,10 +30,11 @@
#define HW_ATL_MPI_CONTROL_ADR 0x0368U
#define HW_ATL_MPI_STATE_ADR 0x036CU
-#define HW_ATL_MPI_STATE_MSK 0x00FFU
-#define HW_ATL_MPI_STATE_SHIFT 0U
-#define HW_ATL_MPI_SPEED_MSK 0xFFFF0000U
-#define HW_ATL_MPI_SPEED_SHIFT 16U
+#define HW_ATL_MPI_STATE_MSK 0x00FFU
+#define HW_ATL_MPI_STATE_SHIFT 0U
+#define HW_ATL_MPI_SPEED_MSK 0x00FF0000U
+#define HW_ATL_MPI_SPEED_SHIFT 16U
+#define HW_ATL_MPI_DIRTY_WAKE_MSK 0x02000000U
#define HW_ATL_MPI_DAISY_CHAIN_STATUS 0x704
#define HW_ATL_MPI_BOOT_EXIT_CODE 0x388
@@ -521,23 +522,24 @@ void hw_atl_utils_mpi_read_stats(struct aq_hw_s *self,
err_exit:;
}
-static int hw_atl_utils_mpi_set_speed(struct aq_hw_s *self, u32 speed)
+int hw_atl_utils_mpi_set_speed(struct aq_hw_s *self, u32 speed)
{
u32 val = aq_hw_read_reg(self, HW_ATL_MPI_CONTROL_ADR);
- val = (val & HW_ATL_MPI_STATE_MSK) | (speed << HW_ATL_MPI_SPEED_SHIFT);
+ val = val & ~HW_ATL_MPI_SPEED_MSK;
+ val |= speed << HW_ATL_MPI_SPEED_SHIFT;
aq_hw_write_reg(self, HW_ATL_MPI_CONTROL_ADR, val);
return 0;
}
-void hw_atl_utils_mpi_set(struct aq_hw_s *self,
- enum hal_atl_utils_fw_state_e state,
- u32 speed)
+int hw_atl_utils_mpi_set_state(struct aq_hw_s *self,
+ enum hal_atl_utils_fw_state_e state)
{
int err = 0;
u32 transaction_id = 0;
struct hw_aq_atl_utils_mbox_header mbox;
+ u32 val = aq_hw_read_reg(self, HW_ATL_MPI_CONTROL_ADR);
if (state == MPI_RESET) {
hw_atl_utils_mpi_read_mbox(self, &mbox);
@@ -551,21 +553,21 @@ void hw_atl_utils_mpi_set(struct aq_hw_s *self,
if (err < 0)
goto err_exit;
}
+ /* On interface DEINIT we disable DW (raise bit)
+ * Otherwise enable DW (clear bit)
+ */
+ if (state == MPI_DEINIT || state == MPI_POWER)
+ val |= HW_ATL_MPI_DIRTY_WAKE_MSK;
+ else
+ val &= ~HW_ATL_MPI_DIRTY_WAKE_MSK;
- aq_hw_write_reg(self, HW_ATL_MPI_CONTROL_ADR,
- (speed << HW_ATL_MPI_SPEED_SHIFT) | state);
+ /* Set new state bits */
+ val = val & ~HW_ATL_MPI_STATE_MSK;
+ val |= state & HW_ATL_MPI_STATE_MSK;
-err_exit:;
-}
-
-static int hw_atl_utils_mpi_set_state(struct aq_hw_s *self,
- enum hal_atl_utils_fw_state_e state)
-{
- u32 val = aq_hw_read_reg(self, HW_ATL_MPI_CONTROL_ADR);
-
- val = state | (val & HW_ATL_MPI_SPEED_MSK);
aq_hw_write_reg(self, HW_ATL_MPI_CONTROL_ADR, val);
- return 0;
+err_exit:
+ return err;
}
int hw_atl_utils_mpi_get_link_status(struct aq_hw_s *self)
@@ -721,16 +723,18 @@ void hw_atl_utils_hw_chip_features_init(struct aq_hw_s *self, u32 *p)
*p = chip_features;
}
-int hw_atl_utils_hw_deinit(struct aq_hw_s *self)
+static int hw_atl_fw1x_deinit(struct aq_hw_s *self)
{
- hw_atl_utils_mpi_set(self, MPI_DEINIT, 0x0U);
+ hw_atl_utils_mpi_set_speed(self, 0);
+ hw_atl_utils_mpi_set_state(self, MPI_DEINIT);
return 0;
}
int hw_atl_utils_hw_set_power(struct aq_hw_s *self,
unsigned int power_state)
{
- hw_atl_utils_mpi_set(self, MPI_POWER, 0x0U);
+ hw_atl_utils_mpi_set_speed(self, 0);
+ hw_atl_utils_mpi_set_state(self, MPI_POWER);
return 0;
}
@@ -823,6 +827,7 @@ int hw_atl_utils_get_fw_version(struct aq_hw_s *self, u32 *fw_version)
const struct aq_fw_ops aq_fw_1x_ops = {
.init = hw_atl_utils_mpi_create,
+ .deinit = hw_atl_fw1x_deinit,
.reset = NULL,
.get_mac_permanent = hw_atl_utils_get_mac_permanent,
.set_link_speed = hw_atl_utils_mpi_set_speed,
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
index 8cfce95..a4ac592 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c
@@ -28,6 +28,10 @@
#define HW_ATL_FW2X_MPI_STATE_ADDR 0x370
#define HW_ATL_FW2X_MPI_STATE2_ADDR 0x374
+static int aq_fw2x_set_link_speed(struct aq_hw_s *self, u32 speed);
+static int aq_fw2x_set_state(struct aq_hw_s *self,
+ enum hal_atl_utils_fw_state_e state);
+
static int aq_fw2x_init(struct aq_hw_s *self)
{
int err = 0;
@@ -39,6 +43,16 @@ static int aq_fw2x_init(struct aq_hw_s *self)
return err;
}
+static int aq_fw2x_deinit(struct aq_hw_s *self)
+{
+ int err = aq_fw2x_set_link_speed(self, 0);
+
+ if (!err)
+ err = aq_fw2x_set_state(self, MPI_DEINIT);
+
+ return err;
+}
+
static enum hw_atl_fw2x_rate link_speed_mask_2fw2x_ratemask(u32 speed)
{
enum hw_atl_fw2x_rate rate = 0;
@@ -76,7 +90,21 @@ static int aq_fw2x_set_link_speed(struct aq_hw_s *self, u32 speed)
static int aq_fw2x_set_state(struct aq_hw_s *self,
enum hal_atl_utils_fw_state_e state)
{
- /* No explicit state in 2x fw */
+ u32 mpi_state = aq_hw_read_reg(self, HW_ATL_FW2X_MPI_CONTROL2_ADDR);
+
+ switch (state) {
+ case MPI_INIT:
+ mpi_state &= ~BIT(CAPS_HI_LINK_DROP);
+ break;
+ case MPI_DEINIT:
+ mpi_state |= BIT(CAPS_HI_LINK_DROP);
+ break;
+ case MPI_RESET:
+ case MPI_POWER:
+ /* No actions */
+ break;
+ }
+ aq_hw_write_reg(self, HW_ATL_FW2X_MPI_CONTROL2_ADDR, mpi_state);
return 0;
}
@@ -175,6 +203,7 @@ static int aq_fw2x_update_stats(struct aq_hw_s *self)
const struct aq_fw_ops aq_fw_2x_ops = {
.init = aq_fw2x_init,
+ .deinit = aq_fw2x_deinit,
.reset = NULL,
.get_mac_permanent = aq_fw2x_get_mac_permanent,
.set_link_speed = aq_fw2x_set_link_speed,
--
2.7.4
^ permalink raw reply related
* [PATCH net-next v2 1/5] net: aquantia: Ethtool based ring size configuration
From: Igor Russkikh @ 2018-06-04 22:30 UTC (permalink / raw)
To: David S . Miller
Cc: netdev, David Arcari, Pavel Belous, Igor Russkikh, Anton Mikaev
In-Reply-To: <cover.1528150073.git.igor.russkikh@aquantia.com>
From: Anton Mikaev <amikaev@aquantia.com>
Implemented ring size setup, min/max validation and reconfiguration in
runtime. NIC level lock is used to prevent collisions on parallel
reconfiguration and interference with periodic service timer job.
Signed-off-by: Anton Mikaev <amikaev@aquantia.com>
Signed-off-by: Igor Russkikh <igor.russkikh@aquantia.com>
---
.../net/ethernet/aquantia/atlantic/aq_ethtool.c | 65 ++++++++++++++++++++++
drivers/net/ethernet/aquantia/atlantic/aq_hw.h | 9 ++-
drivers/net/ethernet/aquantia/atlantic/aq_nic.c | 9 ++-
drivers/net/ethernet/aquantia/atlantic/aq_nic.h | 2 +
.../ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c | 46 +++++++--------
.../aquantia/atlantic/hw_atl/hw_atl_a0_internal.h | 8 +++
.../ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c | 50 +++++++++--------
.../aquantia/atlantic/hw_atl/hw_atl_b0_internal.h | 8 +++
8 files changed, 147 insertions(+), 50 deletions(-)
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c b/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
index f2d8063..97f42d1 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_ethtool.c
@@ -11,6 +11,7 @@
#include "aq_ethtool.h"
#include "aq_nic.h"
+#include "aq_vec.h"
static void aq_ethtool_get_regs(struct net_device *ndev,
struct ethtool_regs *regs, void *p)
@@ -284,6 +285,68 @@ static int aq_ethtool_set_coalesce(struct net_device *ndev,
return aq_nic_update_interrupt_moderation_settings(aq_nic);
}
+static void aq_get_ringparam(struct net_device *ndev,
+ struct ethtool_ringparam *ring)
+{
+ struct aq_nic_s *aq_nic = netdev_priv(ndev);
+ struct aq_nic_cfg_s *aq_nic_cfg = aq_nic_get_cfg(aq_nic);
+
+ ring->rx_pending = aq_nic_cfg->rxds;
+ ring->tx_pending = aq_nic_cfg->txds;
+
+ ring->rx_max_pending = aq_nic_cfg->aq_hw_caps->rxds_max;
+ ring->tx_max_pending = aq_nic_cfg->aq_hw_caps->txds_max;
+}
+
+static int aq_set_ringparam(struct net_device *ndev,
+ struct ethtool_ringparam *ring)
+{
+ int err = 0;
+ bool ndev_running = false;
+ struct aq_nic_s *aq_nic = netdev_priv(ndev);
+ struct aq_nic_cfg_s *aq_nic_cfg = aq_nic_get_cfg(aq_nic);
+ const struct aq_hw_caps_s *hw_caps = aq_nic_cfg->aq_hw_caps;
+
+ if (ring->rx_mini_pending || ring->rx_jumbo_pending) {
+ err = -EOPNOTSUPP;
+ goto err_exit;
+ }
+
+ mutex_lock(&aq_nic->aq_mutex);
+
+ if (netif_running(ndev)) {
+ ndev_running = true;
+ dev_close(ndev);
+ }
+
+ aq_nic_free_vectors(aq_nic);
+
+ aq_nic_cfg->rxds = max(ring->rx_pending, hw_caps->rxds_min);
+ aq_nic_cfg->rxds = min(aq_nic_cfg->rxds, hw_caps->rxds_max);
+ aq_nic_cfg->rxds = ALIGN(aq_nic_cfg->rxds, AQ_HW_RXD_MULTIPLE);
+
+ aq_nic_cfg->txds = max(ring->tx_pending, hw_caps->txds_min);
+ aq_nic_cfg->txds = min(aq_nic_cfg->txds, hw_caps->txds_max);
+ aq_nic_cfg->txds = ALIGN(aq_nic_cfg->txds, AQ_HW_TXD_MULTIPLE);
+
+ for (aq_nic->aq_vecs = 0; aq_nic->aq_vecs < aq_nic_cfg->vecs;
+ aq_nic->aq_vecs++) {
+ aq_nic->aq_vec[aq_nic->aq_vecs] =
+ aq_vec_alloc(aq_nic, aq_nic->aq_vecs, aq_nic_cfg);
+ if (unlikely(!aq_nic->aq_vec[aq_nic->aq_vecs])) {
+ err = -ENOMEM;
+ goto err_unlock;
+ }
+ }
+ if (ndev_running)
+ err = dev_open(ndev);
+
+err_unlock:
+ mutex_unlock(&aq_nic->aq_mutex);
+err_exit:
+ return err;
+}
+
const struct ethtool_ops aq_ethtool_ops = {
.get_link = aq_ethtool_get_link,
.get_regs_len = aq_ethtool_get_regs_len,
@@ -291,6 +354,8 @@ const struct ethtool_ops aq_ethtool_ops = {
.get_drvinfo = aq_ethtool_get_drvinfo,
.get_strings = aq_ethtool_get_strings,
.get_rxfh_indir_size = aq_ethtool_get_rss_indir_size,
+ .get_ringparam = aq_get_ringparam,
+ .set_ringparam = aq_set_ringparam,
.get_rxfh_key_size = aq_ethtool_get_rss_key_size,
.get_rxfh = aq_ethtool_get_rss,
.get_rxnfc = aq_ethtool_get_rxnfc,
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_hw.h b/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
index a2d416b..904cdfd 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_hw.h
@@ -24,8 +24,10 @@ struct aq_hw_caps_s {
u64 link_speed_msk;
unsigned int hw_priv_flags;
u32 media_type;
- u32 rxds;
- u32 txds;
+ u32 rxds_max;
+ u32 txds_max;
+ u32 rxds_min;
+ u32 txds_min;
u32 txhwb_alignment;
u32 irq_mask;
u32 vecs;
@@ -98,6 +100,9 @@ struct aq_stats_s {
#define AQ_HW_MEDIA_TYPE_TP 1U
#define AQ_HW_MEDIA_TYPE_FIBRE 2U
+#define AQ_HW_TXD_MULTIPLE 8U
+#define AQ_HW_RXD_MULTIPLE 8U
+
struct aq_hw_s {
atomic_t flags;
u8 rbl_enabled:1;
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_nic.c b/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
index 1a1a638..a5ccfde 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_nic.c
@@ -89,8 +89,8 @@ void aq_nic_cfg_start(struct aq_nic_s *self)
aq_nic_rss_init(self, cfg->num_rss_queues);
/*descriptors */
- cfg->rxds = min(cfg->aq_hw_caps->rxds, AQ_CFG_RXDS_DEF);
- cfg->txds = min(cfg->aq_hw_caps->txds, AQ_CFG_TXDS_DEF);
+ cfg->rxds = min(cfg->aq_hw_caps->rxds_max, AQ_CFG_RXDS_DEF);
+ cfg->txds = min(cfg->aq_hw_caps->txds_max, AQ_CFG_TXDS_DEF);
/*rss rings */
cfg->vecs = min(cfg->aq_hw_caps->vecs, AQ_CFG_VECS_DEF);
@@ -158,6 +158,8 @@ static void aq_nic_service_timer_cb(struct timer_list *t)
int ctimer = AQ_CFG_SERVICE_TIMER_INTERVAL;
int err = 0;
+ mutex_lock(&self->aq_mutex);
+
if (aq_utils_obj_test(&self->flags, AQ_NIC_FLAGS_IS_NOT_READY))
goto err_exit;
@@ -175,6 +177,7 @@ static void aq_nic_service_timer_cb(struct timer_list *t)
ctimer = max(ctimer / 2, 1);
err_exit:
+ mutex_unlock(&self->aq_mutex);
mod_timer(&self->service_timer, jiffies + ctimer);
}
@@ -288,6 +291,8 @@ int aq_nic_init(struct aq_nic_s *self)
self->aq_vecs > i; ++i, aq_vec = self->aq_vec[i])
aq_vec_init(aq_vec, self->aq_hw_ops, self->aq_hw);
+ mutex_init(&self->aq_mutex);
+
netif_carrier_off(self->ndev);
err_exit:
diff --git a/drivers/net/ethernet/aquantia/atlantic/aq_nic.h b/drivers/net/ethernet/aquantia/atlantic/aq_nic.h
index faa533a..92dbdf1 100644
--- a/drivers/net/ethernet/aquantia/atlantic/aq_nic.h
+++ b/drivers/net/ethernet/aquantia/atlantic/aq_nic.h
@@ -81,6 +81,8 @@ struct aq_nic_s {
struct pci_dev *pdev;
unsigned int msix_entry_mask;
u32 irqvecs;
+ /* NIC reconfiguration synchronization */
+ struct mutex aq_mutex;
};
static inline struct device *aq_nic_get_dev(struct aq_nic_s *self)
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c
index 67e2f9f..7fd6a7e 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c
@@ -19,29 +19,31 @@
#include "hw_atl_a0_internal.h"
#define DEFAULT_A0_BOARD_BASIC_CAPABILITIES \
- .is_64_dma = true, \
- .msix_irqs = 4U, \
- .irq_mask = ~0U, \
- .vecs = HW_ATL_A0_RSS_MAX, \
- .tcs = HW_ATL_A0_TC_MAX, \
- .rxd_alignment = 1U, \
- .rxd_size = HW_ATL_A0_RXD_SIZE, \
- .rxds = 248U, \
- .txd_alignment = 1U, \
- .txd_size = HW_ATL_A0_TXD_SIZE, \
- .txds = 8U * 1024U, \
- .txhwb_alignment = 4096U, \
- .tx_rings = HW_ATL_A0_TX_RINGS, \
- .rx_rings = HW_ATL_A0_RX_RINGS, \
- .hw_features = NETIF_F_HW_CSUM | \
- NETIF_F_RXHASH | \
- NETIF_F_RXCSUM | \
- NETIF_F_SG | \
- NETIF_F_TSO, \
+ .is_64_dma = true, \
+ .msix_irqs = 4U, \
+ .irq_mask = ~0U, \
+ .vecs = HW_ATL_A0_RSS_MAX, \
+ .tcs = HW_ATL_A0_TC_MAX, \
+ .rxd_alignment = 1U, \
+ .rxd_size = HW_ATL_A0_RXD_SIZE, \
+ .rxds_max = HW_ATL_A0_MAX_RXD, \
+ .rxds_min = HW_ATL_A0_MIN_RXD, \
+ .txd_alignment = 1U, \
+ .txd_size = HW_ATL_A0_TXD_SIZE, \
+ .txds_max = HW_ATL_A0_MAX_TXD, \
+ .txds_min = HW_ATL_A0_MIN_RXD, \
+ .txhwb_alignment = 4096U, \
+ .tx_rings = HW_ATL_A0_TX_RINGS, \
+ .rx_rings = HW_ATL_A0_RX_RINGS, \
+ .hw_features = NETIF_F_HW_CSUM | \
+ NETIF_F_RXHASH | \
+ NETIF_F_RXCSUM | \
+ NETIF_F_SG | \
+ NETIF_F_TSO, \
.hw_priv_flags = IFF_UNICAST_FLT, \
- .flow_control = true, \
- .mtu = HW_ATL_A0_MTU_JUMBO, \
- .mac_regs_count = 88, \
+ .flow_control = true, \
+ .mtu = HW_ATL_A0_MTU_JUMBO, \
+ .mac_regs_count = 88, \
.hw_alive_check_addr = 0x10U
const struct aq_hw_caps_s hw_atl_a0_caps_aqc100 = {
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0_internal.h b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0_internal.h
index 1d88555..3c94cff 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0_internal.h
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_a0_internal.h
@@ -88,4 +88,12 @@
#define HW_ATL_A0_FW_VER_EXPECTED 0x01050006U
+#define HW_ATL_A0_MIN_RXD \
+ (ALIGN(AQ_CFG_SKB_FRAGS_MAX + 1U, AQ_HW_RXD_MULTIPLE))
+#define HW_ATL_A0_MIN_TXD \
+ (ALIGN(AQ_CFG_SKB_FRAGS_MAX + 1U, AQ_HW_TXD_MULTIPLE))
+
+#define HW_ATL_A0_MAX_RXD 8184U
+#define HW_ATL_A0_MAX_TXD 8184U
+
#endif /* HW_ATL_A0_INTERNAL_H */
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c
index 819f6bc..4ea15b9 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c
@@ -20,30 +20,32 @@
#include "hw_atl_llh_internal.h"
#define DEFAULT_B0_BOARD_BASIC_CAPABILITIES \
- .is_64_dma = true, \
- .msix_irqs = 4U, \
- .irq_mask = ~0U, \
- .vecs = HW_ATL_B0_RSS_MAX, \
- .tcs = HW_ATL_B0_TC_MAX, \
- .rxd_alignment = 1U, \
- .rxd_size = HW_ATL_B0_RXD_SIZE, \
- .rxds = 4U * 1024U, \
- .txd_alignment = 1U, \
- .txd_size = HW_ATL_B0_TXD_SIZE, \
- .txds = 8U * 1024U, \
- .txhwb_alignment = 4096U, \
- .tx_rings = HW_ATL_B0_TX_RINGS, \
- .rx_rings = HW_ATL_B0_RX_RINGS, \
- .hw_features = NETIF_F_HW_CSUM | \
- NETIF_F_RXCSUM | \
- NETIF_F_RXHASH | \
- NETIF_F_SG | \
- NETIF_F_TSO | \
- NETIF_F_LRO, \
- .hw_priv_flags = IFF_UNICAST_FLT, \
- .flow_control = true, \
- .mtu = HW_ATL_B0_MTU_JUMBO, \
- .mac_regs_count = 88, \
+ .is_64_dma = true, \
+ .msix_irqs = 4U, \
+ .irq_mask = ~0U, \
+ .vecs = HW_ATL_B0_RSS_MAX, \
+ .tcs = HW_ATL_B0_TC_MAX, \
+ .rxd_alignment = 1U, \
+ .rxd_size = HW_ATL_B0_RXD_SIZE, \
+ .rxds_max = HW_ATL_B0_MAX_RXD, \
+ .rxds_min = HW_ATL_B0_MIN_RXD, \
+ .txd_alignment = 1U, \
+ .txd_size = HW_ATL_B0_TXD_SIZE, \
+ .txds_max = HW_ATL_B0_MAX_TXD, \
+ .txds_min = HW_ATL_B0_MIN_TXD, \
+ .txhwb_alignment = 4096U, \
+ .tx_rings = HW_ATL_B0_TX_RINGS, \
+ .rx_rings = HW_ATL_B0_RX_RINGS, \
+ .hw_features = NETIF_F_HW_CSUM | \
+ NETIF_F_RXCSUM | \
+ NETIF_F_RXHASH | \
+ NETIF_F_SG | \
+ NETIF_F_TSO | \
+ NETIF_F_LRO, \
+ .hw_priv_flags = IFF_UNICAST_FLT, \
+ .flow_control = true, \
+ .mtu = HW_ATL_B0_MTU_JUMBO, \
+ .mac_regs_count = 88, \
.hw_alive_check_addr = 0x10U
const struct aq_hw_caps_s hw_atl_b0_caps_aqc100 = {
diff --git a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0_internal.h b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0_internal.h
index 405d145..28568f5 100644
--- a/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0_internal.h
+++ b/drivers/net/ethernet/aquantia/atlantic/hw_atl/hw_atl_b0_internal.h
@@ -142,6 +142,14 @@
#define HW_ATL_INTR_MODER_MAX 0x1FF
#define HW_ATL_INTR_MODER_MIN 0xFF
+#define HW_ATL_B0_MIN_RXD \
+ (ALIGN(AQ_CFG_SKB_FRAGS_MAX + 1U, AQ_HW_RXD_MULTIPLE))
+#define HW_ATL_B0_MIN_TXD \
+ (ALIGN(AQ_CFG_SKB_FRAGS_MAX + 1U, AQ_HW_TXD_MULTIPLE))
+
+#define HW_ATL_B0_MAX_RXD 8184U
+#define HW_ATL_B0_MAX_TXD 8184U
+
/* HW layer capabilities */
#endif /* HW_ATL_B0_INTERNAL_H */
--
2.7.4
^ permalink raw reply related
* [PATCH net-next v2 0/5] net: aquantia: various ethtool ops implementation
From: Igor Russkikh @ 2018-06-04 22:30 UTC (permalink / raw)
To: David S . Miller; +Cc: netdev, David Arcari, Pavel Belous, Igor Russkikh
In this patchset Anton Mikaev and I added some useful ethtool operations:
- ring size changes
- link renegotioation
- flow control management
The patch also improves init/deinit sequence.
V2 changes:
- using mutex to secure simultaneous dev close/open
- using state var to store/restore dev state
Igor Russkikh (5):
net: aquantia: Ethtool based ring size configuration
net: aquantia: Improve adapter init/deinit logic
net: aquantia: Implement rx/tx flow control ethtools callback
net: aquantia: Add renegotiate ethtool operation support
net: aquantia: bump driver version
.../net/ethernet/aquantia/atlantic/aq_ethtool.c | 121 +++++++++++++++++++++
drivers/net/ethernet/aquantia/atlantic/aq_hw.h | 20 +++-
drivers/net/ethernet/aquantia/atlantic/aq_nic.c | 17 ++-
drivers/net/ethernet/aquantia/atlantic/aq_nic.h | 2 +
.../ethernet/aquantia/atlantic/hw_atl/hw_atl_a0.c | 47 ++++----
.../aquantia/atlantic/hw_atl/hw_atl_a0_internal.h | 8 ++
.../ethernet/aquantia/atlantic/hw_atl/hw_atl_b0.c | 51 ++++-----
.../aquantia/atlantic/hw_atl/hw_atl_b0_internal.h | 8 ++
.../aquantia/atlantic/hw_atl/hw_atl_utils.c | 54 +++++----
.../aquantia/atlantic/hw_atl/hw_atl_utils.h | 35 ++++++
.../aquantia/atlantic/hw_atl/hw_atl_utils_fw2x.c | 69 +++++++++++-
drivers/net/ethernet/aquantia/atlantic/ver.h | 4 +-
12 files changed, 352 insertions(+), 84 deletions(-)
--
2.7.4
^ permalink raw reply
* [PATCH net-next] tcp: refactor tcp_ecn_check_ce to remove sk type cast
From: Yousuk Seung @ 2018-06-04 22:29 UTC (permalink / raw)
To: David Miller
Cc: netdev, Yousuk Seung, Neal Cardwell, Yuchung Cheng, Eric Dumazet
Refactor tcp_ecn_check_ce and __tcp_ecn_check_ce to accept struct sock*
instead of tcp_sock* to clean up type casts. This is a pure refactor
patch.
Signed-off-by: Yousuk Seung <ysseung@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
---
net/ipv4/tcp_input.c | 26 ++++++++++++++------------
1 file changed, 14 insertions(+), 12 deletions(-)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index d5ffb573ca4d..355d3dffd021 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -254,8 +254,10 @@ static void tcp_ecn_withdraw_cwr(struct tcp_sock *tp)
tp->ecn_flags &= ~TCP_ECN_DEMAND_CWR;
}
-static void __tcp_ecn_check_ce(struct tcp_sock *tp, const struct sk_buff *skb)
+static void __tcp_ecn_check_ce(struct sock *sk, const struct sk_buff *skb)
{
+ struct tcp_sock *tp = tcp_sk(sk);
+
switch (TCP_SKB_CB(skb)->ip_dsfield & INET_ECN_MASK) {
case INET_ECN_NOT_ECT:
/* Funny extension: if ECT is not set on a segment,
@@ -263,31 +265,31 @@ static void __tcp_ecn_check_ce(struct tcp_sock *tp, const struct sk_buff *skb)
* it is probably a retransmit.
*/
if (tp->ecn_flags & TCP_ECN_SEEN)
- tcp_enter_quickack_mode((struct sock *)tp, 1);
+ tcp_enter_quickack_mode(sk, 1);
break;
case INET_ECN_CE:
- if (tcp_ca_needs_ecn((struct sock *)tp))
- tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_IS_CE);
+ if (tcp_ca_needs_ecn(sk))
+ tcp_ca_event(sk, CA_EVENT_ECN_IS_CE);
if (!(tp->ecn_flags & TCP_ECN_DEMAND_CWR)) {
/* Better not delay acks, sender can have a very low cwnd */
- tcp_enter_quickack_mode((struct sock *)tp, 1);
+ tcp_enter_quickack_mode(sk, 1);
tp->ecn_flags |= TCP_ECN_DEMAND_CWR;
}
tp->ecn_flags |= TCP_ECN_SEEN;
break;
default:
- if (tcp_ca_needs_ecn((struct sock *)tp))
- tcp_ca_event((struct sock *)tp, CA_EVENT_ECN_NO_CE);
+ if (tcp_ca_needs_ecn(sk))
+ tcp_ca_event(sk, CA_EVENT_ECN_NO_CE);
tp->ecn_flags |= TCP_ECN_SEEN;
break;
}
}
-static void tcp_ecn_check_ce(struct tcp_sock *tp, const struct sk_buff *skb)
+static void tcp_ecn_check_ce(struct sock *sk, const struct sk_buff *skb)
{
- if (tp->ecn_flags & TCP_ECN_OK)
- __tcp_ecn_check_ce(tp, skb);
+ if (tcp_sk(sk)->ecn_flags & TCP_ECN_OK)
+ __tcp_ecn_check_ce(sk, skb);
}
static void tcp_ecn_rcv_synack(struct tcp_sock *tp, const struct tcphdr *th)
@@ -710,7 +712,7 @@ static void tcp_event_data_recv(struct sock *sk, struct sk_buff *skb)
}
icsk->icsk_ack.lrcvtime = now;
- tcp_ecn_check_ce(tp, skb);
+ tcp_ecn_check_ce(sk, skb);
if (skb->len >= 128)
tcp_grow_window(sk, skb);
@@ -4434,7 +4436,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
u32 seq, end_seq;
bool fragstolen;
- tcp_ecn_check_ce(tp, skb);
+ tcp_ecn_check_ce(sk, skb);
if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) {
NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFODROP);
--
2.17.1.1185.g55be947832-goog
^ permalink raw reply related
* Re: [PATCH bpf-next v2 1/2] trace_helpers.c: Add helpers to poll multiple perf FDs for events
From: Jakub Kicinski @ 2018-06-04 22:26 UTC (permalink / raw)
To: Toke Høiland-Jørgensen
Cc: netdev, Daniel Borkmann, Alexei Starovoitov
In-Reply-To: <152813003609.3465.618891361534945522.stgit@alrua-kau>
On Mon, 04 Jun 2018 18:33:56 +0200, Toke Høiland-Jørgensen wrote:
> This adds two new helper functions to trace_helpers that supports polling
> multiple perf file descriptors for events. These are used to the XDP
> perf_event_output example, which needs to work with one perf fd per CPU.
>
> Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
Did you take a look at tools/bpf/bpftool/map_perf_ring.c ?
I think the ability to poll multiple FDs could be generally useful and
therefore better add it to libbpf.c than
tools/testing/selftests/bpf/trace_helpers.c? I'm not 100% sure myself...
^ permalink raw reply
* [PATCH net-next v11 10/10] MAINTAINERS: Add entry for drivers/net/ethernet/cavium/octeon/octeon3-*
From: Steven J. Hill @ 2018-06-04 22:00 UTC (permalink / raw)
To: netdev; +Cc: David Daney
In-Reply-To: <1528149617-8964-1-git-send-email-steven.hill@cavium.com>
From: David Daney <david.daney@cavium.com>
Signed-off-by: David Daney <david.daney@cavium.com>
---
MAINTAINERS | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/MAINTAINERS b/MAINTAINERS
index 70d61c2..9ab8b69 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -3249,6 +3249,12 @@ W: http://www.cavium.com
S: Supported
F: drivers/mmc/host/cavium*
+CAVIUM OCTEON-III NETWORK DRIVER
+M: Steven J. Hill <Steven.Hill@cavium.com>
+L: netdev@vger.kernel.org
+S: Supported
+F: drivers/net/ethernet/cavium/octeon/octeon3-*
+
CAVIUM OCTEON-TX CRYPTO DRIVER
M: George Cherian <george.cherian@cavium.com>
L: linux-crypto@vger.kernel.org
--
2.1.4
^ permalink raw reply related
* [PATCH net-next v11 09/10] netdev: cavium: octeon: Add Octeon III BGX Ethernet building
From: Steven J. Hill @ 2018-06-04 22:00 UTC (permalink / raw)
To: netdev; +Cc: Carlos Munoz, Steven J. Hill
In-Reply-To: <1528149617-8964-1-git-send-email-steven.hill@cavium.com>
From: Carlos Munoz <cmunoz@cavium.com>
Add the build and configuration files for the BGX Ethernet.
Signed-off-by: Carlos Munoz <cmunoz@cavium.com>
Signed-off-by: Steven J. Hill <Steven.Hill@cavium.com>
---
drivers/net/ethernet/cavium/Kconfig | 22 +++++++++++++++++++++-
drivers/net/ethernet/cavium/octeon/Makefile | 8 +++++++-
2 files changed, 28 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/cavium/Kconfig b/drivers/net/ethernet/cavium/Kconfig
index 043e3c1..3b9709d 100644
--- a/drivers/net/ethernet/cavium/Kconfig
+++ b/drivers/net/ethernet/cavium/Kconfig
@@ -4,7 +4,7 @@
config NET_VENDOR_CAVIUM
bool "Cavium ethernet drivers"
- depends on PCI
+ depends on PCI || CAVIUM_OCTEON_SOC
default y
---help---
Select this option if you want enable Cavium network support.
@@ -100,4 +100,24 @@ config LIQUIDIO_VF
will be called liquidio_vf. MSI-X interrupt support is required
for this driver to work correctly
+config OCTEON3_BGX_PORT
+ tristate "Cavium Octeon III BGX port support"
+ depends on CAVIUM_OCTEON_SOC
+ ---help---
+ This driver adds support for Cavium Octeon III BGX ports. BGX ports
+ support sgmii, rgmii, xaui, rxaui, xlaui, xfi, 10KR and 40KR modes.
+
+ Say Y to use the management port on Octeon III boards or to use
+ any other ethernet port.
+
+config OCTEON3_ETHERNET
+ tristate "Cavium OCTEON III PKI/PKO Ethernet support"
+ depends on CAVIUM_OCTEON_SOC
+ select OCTEON_BGX_PORT
+ select OCTEON_FPA3
+ select FW_LOADER
+ ---help---
+ Support for 'BGX' Ethernet via PKI/PKO units. No support for
+ cn70xx chips, use OCTEON_ETHERNET instead.
+
endif # NET_VENDOR_CAVIUM
diff --git a/drivers/net/ethernet/cavium/octeon/Makefile b/drivers/net/ethernet/cavium/octeon/Makefile
index efa41c1..1939c84 100644
--- a/drivers/net/ethernet/cavium/octeon/Makefile
+++ b/drivers/net/ethernet/cavium/octeon/Makefile
@@ -1,5 +1,11 @@
+# SPDX-License-Identifier: GPL-2.0
#
# Makefile for the Cavium network device drivers.
#
-obj-$(CONFIG_OCTEON_MGMT_ETHERNET) += octeon_mgmt.o
+obj-$(CONFIG_OCTEON_MGMT_ETHERNET) += octeon_mgmt.o
+obj-$(CONFIG_OCTEON3_BGX_PORT) += octeon3-bgx-nexus.o octeon3-bgx-port.o
+obj-$(CONFIG_OCTEON3_ETHERNET) += octeon3-ethernet.o
+
+octeon3-ethernet-objs += octeon3-core.o octeon3-pki.o octeon3-pko.o \
+ octeon3-sso.o
--
2.1.4
^ permalink raw reply related
* [PATCH net-next v11 08/10] netdev: cavium: octeon: Add Octeon III BGX Ethernet core
From: Steven J. Hill @ 2018-06-04 22:00 UTC (permalink / raw)
To: netdev; +Cc: Carlos Munoz, Steven J. Hill
In-Reply-To: <1528149617-8964-1-git-send-email-steven.hill@cavium.com>
From: Carlos Munoz <cmunoz@cavium.com>
This is the main core of the BGX Ethernet driver.
Signed-off-by: Carlos Munoz <cmunoz@cavium.com>
Signed-off-by: Steven J. Hill <Steven.Hill@cavium.com>
---
drivers/net/ethernet/cavium/octeon/octeon3-core.c | 2380 +++++++++++++++++++++
1 file changed, 2380 insertions(+)
create mode 100644 drivers/net/ethernet/cavium/octeon/octeon3-core.c
diff --git a/drivers/net/ethernet/cavium/octeon/octeon3-core.c b/drivers/net/ethernet/cavium/octeon/octeon3-core.c
new file mode 100644
index 0000000..1e2f68d
--- /dev/null
+++ b/drivers/net/ethernet/cavium/octeon/octeon3-core.c
@@ -0,0 +1,2380 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Octeon III BGX Nexus Ethernet driver core
+ *
+ * Copyright (C) 2018 Cavium, Inc.
+ */
+#include <linux/etherdevice.h>
+#include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/kthread.h>
+#include <linux/net_tstamp.h>
+#include <linux/ptp_clock_kernel.h>
+#include <linux/timecounter.h>
+
+#include "octeon3.h"
+
+/* First buffer:
+ *
+ * +---SKB---------+
+ * | |
+ * | |
+ * +--+--*data |
+ * | | |
+ * | | |
+ * | +---------------+
+ * | /|\
+ * | |
+ * | |
+ * \|/ |
+ * WQE - 128 -+-----> +-------------+-------+ -+-
+ * | | *skb ----+ | |
+ * | | | |
+ * | | | |
+ * WQE_SKIP = 128 | | |
+ * | | | |
+ * | | | |
+ * | | | |
+ * | | | First Skip
+ * WQE -----+-----> +---------------------+ |
+ * | word 0 | |
+ * | word 1 | |
+ * | word 2 | |
+ * | word 3 | |
+ * | word 4 | |
+ * +---------------------+ -+-
+ * +----+- packet link |
+ * | | packet data |
+ * | | |
+ * | | |
+ * | | . |
+ * | | . |
+ * | | . |
+ * | +---------------------+
+ * |
+ * |
+ * Later buffers:|
+ * |
+ * |
+ * |
+ * |
+ * |
+ * | +---SKB---------+
+ * | | |
+ * | | |
+ * | +--+--*data |
+ * | | | |
+ * | | | |
+ * | | +---------------+
+ * | | /|\
+ * | | |
+ * | | |
+ * | \|/ |
+ * WQE - 128 ----+--> +-------------+-------+ -+-
+ * | | *skb ----+ | |
+ * | | | |
+ * | | | |
+ * | | | |
+ * | | | LATER_SKIP = 128
+ * | | | |
+ * | | | |
+ * | | | |
+ * | +---------------------+ -+-
+ * | | packet link |
+ * +--> | packet data |
+ * | |
+ * | |
+ * | . |
+ * | . |
+ * | . |
+ * +---------------------+
+ */
+
+#define MAX_TX_QUEUE_DEPTH 512
+#define SSO_INTSN_EXE 0x61
+#define MAX_RX_CONTEXTS 32
+
+#define SKB_PTR_OFFSET 0
+#define SKB_AURA_OFFSET 1
+#define SKB_AURA_MAGIC 0xbadc0ffee4dad000ULL
+
+#define MAX_CORES 48
+#define FPA3_NUM_AURAS 1024
+
+#define USE_ASYNC_IOBDMA 1
+#define SCR_SCRATCH 0ull
+#define SSO_NO_WAIT 0ull
+#define DID_TAG_SWTAG 0x60ull
+#define IOBDMA_SENDSINGLE 0xffffffffffffa200ull
+
+/* Values for the value of wqe word2 [ERRLEV] */
+#define PKI_ERRLEV_LA 0x01
+
+/* Values for the value of wqe word2 [OPCODE] */
+#define PKI_OPCODE_NONE 0x00
+#define PKI_OPCODE_JABBER 0x02
+#define PKI_OPCODE_FCS 0x07
+
+/* Values for the layer type in the wqe */
+#define PKI_LTYPE_IP4 0x08
+#define PKI_LTYPE_IP6 0x0a
+#define PKI_LTYPE_TCP 0x10
+#define PKI_LTYPE_UDP 0x11
+#define PKI_LTYPE_SCTP 0x12
+
+/* Registers are accessed via xkphys */
+#define SSO_BASE 0x1670000000000ull
+#define SSO_ADDR(node) (SET_XKPHYS + NODE_OFFSET(node) + \
+ SSO_BASE)
+#define GRP_OFFSET(grp) ((grp) << 16)
+#define GRP_ADDR(n, g) (SSO_ADDR(n) + GRP_OFFSET(g))
+#define SSO_GRP_AQ_CNT(n, g) (GRP_ADDR(n, g) + 0x20000700)
+
+#define MIO_PTP_BASE 0x1070000000000ull
+#define MIO_PTP_ADDR(node) (SET_XKPHYS + NODE_OFFSET(node) + \
+ MIO_PTP_BASE)
+#define MIO_PTP_CLOCK_CFG(node) (MIO_PTP_ADDR(node) + 0xf00)
+#define MIO_PTP_CLOCK_HI(node) (MIO_PTP_ADDR(node) + 0xf10)
+#define MIO_PTP_CLOCK_COMP(node) (MIO_PTP_ADDR(node) + 0xf18)
+
+struct octeon3_napi_wrapper {
+ struct napi_struct napi;
+ int available;
+ int idx;
+ int cpu;
+ struct octeon3_rx *cxt;
+} ____cacheline_aligned_in_smp;
+
+/* Up to 2 napis per core are supported */
+#define MAX_NAPI_PER_CPU 2
+#define MAX_NAPIS_PER_NODE (MAX_CORES * MAX_NAPI_PER_CPU)
+
+static struct octeon3_napi_wrapper
+napi_wrapper[MAX_NODES][MAX_NAPIS_PER_NODE]
+__cacheline_aligned_in_smp;
+
+struct octeon3_ethernet;
+
+struct octeon3_rx {
+ struct octeon3_napi_wrapper *napiw;
+ DECLARE_BITMAP(napi_idx_bitmap, MAX_CORES);
+ spinlock_t napi_idx_lock; /* Protect the napi index bitmap */
+ struct octeon3_ethernet *parent;
+ int rx_grp;
+ int rx_irq;
+ cpumask_t rx_affinity_hint;
+};
+
+struct octeon3_ethernet {
+ struct bgx_port_netdev_priv bgx_priv; /* Must be first element. */
+ struct list_head list;
+ struct net_device *netdev;
+ enum octeon3_mac_type mac_type;
+ struct octeon3_rx rx_cxt[MAX_RX_CONTEXTS];
+ struct ptp_clock_info ptp_info;
+ struct ptp_clock *ptp_clock;
+ struct cyclecounter cc;
+ struct timecounter tc;
+ spinlock_t ptp_lock; /* Serialize ptp clock adjustments */
+ int num_rx_cxt;
+ int pki_aura;
+ int pknd;
+ int pko_queue;
+ int node;
+ int interface;
+ int index;
+ int rx_buf_count;
+ int tx_complete_grp;
+ int rx_timestamp_hw:1;
+ int tx_timestamp_hw:1;
+ spinlock_t stat_lock; /* Protects stats counters */
+ u64 last_packets;
+ u64 last_octets;
+ u64 last_dropped;
+ atomic64_t rx_packets;
+ atomic64_t rx_octets;
+ atomic64_t rx_dropped;
+ atomic64_t rx_errors;
+ atomic64_t rx_length_errors;
+ atomic64_t rx_crc_errors;
+ atomic64_t tx_packets;
+ atomic64_t tx_octets;
+ atomic64_t tx_dropped;
+ /* The following two fields need to be on a different cache line as
+ * they are updated by pko which invalidates the cache every time it
+ * updates them. The idea is to prevent other fields from being
+ * invalidated unnecessarily.
+ */
+ char cacheline_pad1[CVMX_CACHE_LINE_SIZE];
+ atomic64_t buffers_needed;
+ atomic64_t tx_backlog;
+ char cacheline_pad2[CVMX_CACHE_LINE_SIZE];
+};
+
+static DEFINE_MUTEX(octeon3_eth_init_mutex);
+
+struct octeon3_ethernet_node;
+
+struct octeon3_ethernet_worker {
+ wait_queue_head_t queue;
+ struct task_struct *task;
+ struct octeon3_ethernet_node *oen;
+ atomic_t kick;
+ int order;
+};
+
+struct octeon3_ethernet_node {
+ bool init_done;
+ bool napi_init_done;
+ int next_cpu_irq_affinity;
+ int node;
+ int pki_packet_pool;
+ int sso_pool;
+ int pko_pool;
+ void *sso_pool_stack;
+ void *pko_pool_stack;
+ void *pki_packet_pool_stack;
+ int sso_aura;
+ int pko_aura;
+ int tx_complete_grp;
+ int tx_irq;
+ cpumask_t tx_affinity_hint;
+ struct octeon3_ethernet_worker workers[8];
+ struct mutex device_list_lock; /* Protects the device list */
+ struct list_head device_list;
+ spinlock_t napi_alloc_lock; /* Protects napi allocations */
+};
+
+/* This array keeps track of the number of napis running on each cpu */
+static u8 octeon3_cpu_napi_cnt[NR_CPUS] __cacheline_aligned_in_smp;
+
+static int use_tx_queues;
+module_param(use_tx_queues, int, 0644);
+MODULE_PARM_DESC(use_tx_queues, "Use network layer transmit queues.");
+
+static int wait_pko_response;
+module_param(wait_pko_response, int, 0644);
+MODULE_PARM_DESC(wait_pko_response, "Wait for response after each pko command.");
+
+static int num_packet_buffers = 768;
+module_param(num_packet_buffers, int, 0444);
+MODULE_PARM_DESC(num_packet_buffers, "Number of packet buffers to allocate per port.");
+
+static int packet_buffer_size = 2048;
+module_param(packet_buffer_size, int, 0444);
+MODULE_PARM_DESC(packet_buffer_size, "Size of each RX packet buffer.");
+
+static int rx_contexts = 1;
+module_param(rx_contexts, int, 0444);
+MODULE_PARM_DESC(rx_contexts, "Number of RX threads per port.");
+
+int ilk0_lanes = 1;
+module_param(ilk0_lanes, int, 0444);
+MODULE_PARM_DESC(ilk0_lanes, "Number of SerDes lanes used by ILK link 0.");
+
+int ilk1_lanes = 1;
+module_param(ilk1_lanes, int, 0444);
+MODULE_PARM_DESC(ilk1_lanes, "Number of SerDes lanes used by ILK link 1.");
+
+static struct octeon3_ethernet_node octeon3_eth_node[MAX_NODES];
+static struct kmem_cache *octeon3_eth_sso_pko_cache;
+
+/* Reads a 64 bit value from the processor local scratchpad memory.
+ *
+ * @param offset byte offset into scratch pad to read
+ *
+ * @return value read
+ */
+static inline u64 scratch_read64(u64 offset)
+{
+ return *(u64 *)((long)SCRATCH_BASE + offset);
+}
+
+/* Write a 64 bit value to the processor local scratchpad memory.
+ *
+ * @param offset byte offset into scratch pad to write
+ @ @praram value to write
+ */
+static inline void scratch_write64(u64 offset, u64 value)
+{
+ *(u64 *)((long)SCRATCH_BASE + offset) = value;
+}
+
+static int get_pki_chan(int node, int interface, int index)
+{
+ int pki_chan;
+
+ pki_chan = node << 12;
+
+ if (OCTEON_IS_MODEL(OCTEON_CNF75XX) &&
+ (interface == 1 || interface == 2)) {
+ /* SRIO */
+ pki_chan |= 0x240 + (2 * (interface - 1)) + index;
+ } else {
+ /* BGX */
+ pki_chan |= 0x800 + (0x100 * interface) + (0x10 * index);
+ }
+
+ return pki_chan;
+}
+
+/* Map auras to the field priv->buffers_needed. Used to speed up packet
+ * transmission.
+ */
+static void *aura2bufs_needed[MAX_NODES][FPA3_NUM_AURAS];
+
+static int octeon3_eth_lgrp_to_ggrp(int node, int grp)
+{
+ return (node << 8) | grp;
+}
+
+static void octeon3_eth_gen_affinity(int node, cpumask_t *mask)
+{
+ int cpu;
+
+ do {
+ cpu = cpumask_next(octeon3_eth_node[node].next_cpu_irq_affinity,
+ cpu_online_mask);
+ octeon3_eth_node[node].next_cpu_irq_affinity++;
+ if (cpu >= nr_cpu_ids) {
+ octeon3_eth_node[node].next_cpu_irq_affinity = -1;
+ continue;
+ }
+ } while (false);
+ cpumask_clear(mask);
+ cpumask_set_cpu(cpu, mask);
+}
+
+struct wr_ret {
+ void *work;
+ u16 grp;
+};
+
+static inline struct wr_ret octeon3_core_get_work_sync(int grp)
+{
+ u64 node = cvmx_get_node_num();
+ u64 addr, response;
+ struct wr_ret r;
+
+ /* See SSO_GET_WORK_LD_S for the address to read */
+ addr = 1ull << 63;
+ addr |= BIT(48);
+ addr |= DID_TAG_SWTAG << 40;
+ addr |= node << 36;
+ addr |= BIT(30);
+ addr |= BIT(29);
+ addr |= octeon3_eth_lgrp_to_ggrp(node, grp) << 4;
+ addr |= SSO_NO_WAIT << 3;
+ response = __raw_readq((void __iomem *)addr);
+
+ /* See SSO_GET_WORK_RTN_S for the format of the response */
+ r.grp = (response & GENMASK_ULL(57, 48)) >> 48;
+ if (response & BIT(63))
+ r.work = NULL;
+ else
+ r.work = phys_to_virt(response & GENMASK_ULL(41, 0));
+
+ return r;
+}
+
+/* octeon3_core_get_work_async - Request work via a iobdma command. Doesn't wait
+ * for the response.
+ *
+ * @grp: Group to request work for.
+ */
+static inline void octeon3_core_get_work_async(unsigned int grp)
+{
+ u64 data, node = cvmx_get_node_num();
+
+ /* See SSO_GET_WORK_DMA_S for the command structure */
+ data = SCR_SCRATCH << 56;
+ data |= 1ull << 48;
+ data |= DID_TAG_SWTAG << 40;
+ data |= node << 36;
+ data |= 1ull << 30;
+ data |= 1ull << 29;
+ data |= octeon3_eth_lgrp_to_ggrp(node, grp) << 4;
+ data |= SSO_NO_WAIT << 3;
+
+ __raw_writeq(data, (void __iomem *)IOBDMA_SENDSINGLE);
+}
+
+/* octeon3_core_get_response_async - Read the request work response. Must be
+ * called after calling
+ * octeon3_core_get_work_async().
+ *
+ * Returns work queue entry.
+ */
+static inline struct wr_ret octeon3_core_get_response_async(void)
+{
+ struct wr_ret r;
+ u64 response;
+
+ CVMX_SYNCIOBDMA;
+ response = scratch_read64(SCR_SCRATCH);
+
+ /* See SSO_GET_WORK_RTN_S for the format of the response */
+ r.grp = (response & GENMASK_ULL(57, 48)) >> 48;
+ if (response & BIT(63))
+ r.work = NULL;
+ else
+ r.work = phys_to_virt(response & GENMASK_ULL(41, 0));
+
+ return r;
+}
+
+static void octeon3_eth_replenish_rx(struct octeon3_ethernet *priv, int count)
+{
+ struct sk_buff *skb;
+ int i;
+
+ for (i = 0; i < count; i++) {
+ void **buf;
+
+ skb = __alloc_skb(packet_buffer_size, GFP_ATOMIC, 0,
+ priv->node);
+ if (!skb)
+ break;
+ buf = (void **)PTR_ALIGN(skb->head, 128);
+ buf[SKB_PTR_OFFSET] = skb;
+ octeon_fpa3_free(priv->node, priv->pki_aura, buf);
+ }
+}
+
+static bool octeon3_eth_tx_done_runnable(struct octeon3_ethernet_worker *worker)
+{
+ return atomic_read(&worker->kick) != 0 || kthread_should_stop();
+}
+
+static int octeon3_eth_replenish_all(struct octeon3_ethernet_node *oen)
+{
+ int batch_size = 32, pending = 0;
+ struct octeon3_ethernet *priv;
+
+ rcu_read_lock();
+ list_for_each_entry_rcu(priv, &oen->device_list, list) {
+ int amount = atomic64_sub_if_positive(batch_size,
+ &priv->buffers_needed);
+
+ if (amount >= 0) {
+ octeon3_eth_replenish_rx(priv, batch_size);
+ pending += amount;
+ }
+ }
+ rcu_read_unlock();
+ return pending;
+}
+
+static int octeon3_eth_tx_complete_hwtstamp(struct octeon3_ethernet *priv,
+ struct sk_buff *skb)
+{
+ struct skb_shared_hwtstamps shts;
+ u64 hwts, ns;
+
+ hwts = *((u64 *)(skb->cb) + 1);
+ ns = timecounter_cyc2time(&priv->tc, hwts);
+ memset(&shts, 0, sizeof(shts));
+ shts.hwtstamp = ns_to_ktime(ns);
+ skb_tstamp_tx(skb, &shts);
+
+ return 0;
+}
+
+static int octeon3_eth_tx_complete_worker(void *data)
+{
+ int backlog, backlog_stop_thresh, i, order, tx_complete_stop_thresh;
+ struct octeon3_ethernet_worker *worker = data;
+ struct octeon3_ethernet_node *oen = worker->oen;
+ u64 aq_cnt;
+
+ order = worker->order;
+ backlog_stop_thresh = (order == 0 ? 31 : order * 80);
+ tx_complete_stop_thresh = (order * 100);
+
+ while (!kthread_should_stop()) {
+ /* Replaced by wait_event to avoid warnings like
+ * "task oct3_eth/0:2:1250 blocked for more than 120 seconds."
+ */
+ wait_event_interruptible(worker->queue,
+ octeon3_eth_tx_done_runnable(worker));
+ atomic_dec_if_positive(&worker->kick); /* clear the flag */
+
+ do {
+ backlog = octeon3_eth_replenish_all(oen);
+ for (i = 0; i < 100; i++) {
+ void **work;
+ struct net_device *tx_netdev;
+ struct octeon3_ethernet *tx_priv;
+ struct sk_buff *skb;
+ struct wr_ret r;
+
+ r = octeon3_core_get_work_sync(oen->tx_complete_grp);
+ work = r.work;
+ if (!work)
+ break;
+ tx_netdev = work[0];
+ tx_priv = netdev_priv(tx_netdev);
+ if (unlikely(netif_queue_stopped(tx_netdev)) && atomic64_read(&tx_priv->tx_backlog) < MAX_TX_QUEUE_DEPTH)
+ netif_wake_queue(tx_netdev);
+ skb = container_of((void *)work,
+ struct sk_buff, cb);
+ if (unlikely(tx_priv->tx_timestamp_hw) && unlikely(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
+ octeon3_eth_tx_complete_hwtstamp(tx_priv, skb);
+ dev_kfree_skb(skb);
+ }
+
+ aq_cnt = oct_csr_read(SSO_GRP_AQ_CNT(oen->node, oen->tx_complete_grp));
+ aq_cnt &= GENMASK_ULL(32, 0);
+ if ((backlog > backlog_stop_thresh ||
+ aq_cnt > tx_complete_stop_thresh) &&
+ order < ARRAY_SIZE(oen->workers) - 1) {
+ atomic_set(&oen->workers[order + 1].kick, 1);
+ wake_up(&oen->workers[order + 1].queue);
+ }
+ } while (!need_resched() && (backlog > backlog_stop_thresh ||
+ aq_cnt > tx_complete_stop_thresh));
+
+ cond_resched();
+
+ if (!octeon3_eth_tx_done_runnable(worker))
+ octeon3_sso_irq_set(oen->node, oen->tx_complete_grp,
+ true);
+ }
+
+ return 0;
+}
+
+static irqreturn_t octeon3_eth_tx_handler(int irq, void *info)
+{
+ struct octeon3_ethernet_node *oen = info;
+
+ /* Disarm the irq. */
+ octeon3_sso_irq_set(oen->node, oen->tx_complete_grp, false);
+ atomic_set(&oen->workers[0].kick, 1);
+ wake_up(&oen->workers[0].queue);
+ return IRQ_HANDLED;
+}
+
+static int octeon3_eth_global_init(unsigned int node,
+ struct platform_device *pdev)
+{
+ struct octeon3_ethernet_node *oen;
+ unsigned int sso_intsn;
+ int i, rv = 0;
+
+ mutex_lock(&octeon3_eth_init_mutex);
+
+ oen = octeon3_eth_node + node;
+
+ if (oen->init_done)
+ goto done;
+
+ /* CN78XX-P1.0 cannot un-initialize PKO, so get a module
+ * reference to prevent it from being unloaded.
+ */
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_0))
+ if (!try_module_get(THIS_MODULE))
+ dev_err(&pdev->dev, "ERROR: Could not obtain module reference for CN78XX-P1.0\n");
+
+ INIT_LIST_HEAD(&oen->device_list);
+ mutex_init(&oen->device_list_lock);
+ spin_lock_init(&oen->napi_alloc_lock);
+
+ oen->node = node;
+
+ octeon_fpa3_init(node);
+ rv = octeon_fpa3_pool_init(node, -1, &oen->sso_pool,
+ &oen->sso_pool_stack, 40960);
+ if (rv)
+ goto done;
+
+ rv = octeon_fpa3_pool_init(node, -1, &oen->pko_pool,
+ &oen->pko_pool_stack, 40960);
+ if (rv)
+ goto done;
+
+ rv = octeon_fpa3_pool_init(node, -1, &oen->pki_packet_pool,
+ &oen->pki_packet_pool_stack,
+ 64 * num_packet_buffers);
+ if (rv)
+ goto done;
+
+ rv = octeon_fpa3_aura_init(node, oen->sso_pool, -1,
+ &oen->sso_aura, num_packet_buffers, 20480);
+ if (rv)
+ goto done;
+
+ rv = octeon_fpa3_aura_init(node, oen->pko_pool, -1,
+ &oen->pko_aura, num_packet_buffers, 20480);
+ if (rv)
+ goto done;
+
+ dev_info(&pdev->dev, "SSO:%d:%d, PKO:%d:%d\n", oen->sso_pool,
+ oen->sso_aura, oen->pko_pool, oen->pko_aura);
+
+ if (!octeon3_eth_sso_pko_cache) {
+ octeon3_eth_sso_pko_cache = kmem_cache_create("sso_pko", 4096,
+ 128, 0, NULL);
+ if (!octeon3_eth_sso_pko_cache) {
+ rv = -ENOMEM;
+ goto done;
+ }
+ }
+
+ rv = octeon_fpa3_mem_fill(node, octeon3_eth_sso_pko_cache,
+ oen->sso_aura, 1024);
+ if (rv)
+ goto done;
+
+ rv = octeon_fpa3_mem_fill(node, octeon3_eth_sso_pko_cache,
+ oen->pko_aura, 1024);
+ if (rv)
+ goto done;
+
+ rv = octeon3_sso_init(node, oen->sso_aura);
+ if (rv)
+ goto done;
+
+ oen->tx_complete_grp = octeon3_sso_alloc_groups(node, NULL, 1, -1);
+ if (oen->tx_complete_grp < 0)
+ goto done;
+
+ sso_intsn = SSO_INTSN_EXE << 12 | oen->tx_complete_grp;
+ oen->tx_irq = irq_create_mapping(NULL, sso_intsn);
+ if (!oen->tx_irq) {
+ rv = -ENODEV;
+ goto done;
+ }
+
+ rv = octeon3_pko_init_global(node, oen->pko_aura);
+ if (rv) {
+ rv = -ENODEV;
+ goto done;
+ }
+
+ octeon3_pki_vlan_init(node);
+ octeon3_pki_cluster_init(node, pdev);
+ octeon3_pki_ltype_init(node);
+ octeon3_pki_enable(node);
+
+ for (i = 0; i < ARRAY_SIZE(oen->workers); i++) {
+ oen->workers[i].oen = oen;
+ init_waitqueue_head(&oen->workers[i].queue);
+ oen->workers[i].order = i;
+ }
+ for (i = 0; i < ARRAY_SIZE(oen->workers); i++) {
+ oen->workers[i].task =
+ kthread_create_on_node(octeon3_eth_tx_complete_worker,
+ oen->workers + i, node,
+ "oct3_eth/%d:%d", node, i);
+ if (IS_ERR(oen->workers[i].task)) {
+ rv = PTR_ERR(oen->workers[i].task);
+ goto done;
+ } else {
+#ifdef CONFIG_NUMA
+ set_cpus_allowed_ptr(oen->workers[i].task,
+ cpumask_of_node(node));
+#endif
+ wake_up_process(oen->workers[i].task);
+ }
+ }
+
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X))
+ octeon3_sso_pass1_limit(node, oen->tx_complete_grp);
+
+ rv = request_irq(oen->tx_irq, octeon3_eth_tx_handler,
+ IRQ_TYPE_EDGE_RISING, "oct3_eth_tx_done", oen);
+ if (rv)
+ goto done;
+ octeon3_eth_gen_affinity(node, &oen->tx_affinity_hint);
+ irq_set_affinity_hint(oen->tx_irq, &oen->tx_affinity_hint);
+
+ octeon3_sso_irq_set(node, oen->tx_complete_grp, true);
+
+ oen->init_done = true;
+done:
+ mutex_unlock(&octeon3_eth_init_mutex);
+ return rv;
+}
+
+static struct sk_buff *octeon3_eth_work_to_skb(void *w)
+{
+ struct sk_buff *skb;
+ void **f = w;
+
+ skb = f[-16];
+ return skb;
+}
+
+/* octeon3_napi_alloc_cpu - Find an available cpu. This function must be called
+ * with the napi_alloc_lock lock held.
+ * @node: Node to allocate cpu from.
+ * @cpu: Cpu to bind the napi to:
+ * < 0: use any cpu.
+ * >= 0: use requested cpu.
+ *
+ * Returns cpu number.
+ * Returns <0 for error codes.
+ */
+static int octeon3_napi_alloc_cpu(int node, int cpu)
+{
+ int min_cnt = MAX_NAPIS_PER_NODE;
+ int min_cpu = -EBUSY;
+
+ if (cpu >= 0) {
+ min_cpu = cpu;
+ } else {
+ for_each_cpu(cpu, cpumask_of_node(node)) {
+ if (octeon3_cpu_napi_cnt[cpu] == 0) {
+ min_cpu = cpu;
+ break;
+ } else if (octeon3_cpu_napi_cnt[cpu] < min_cnt) {
+ min_cnt = octeon3_cpu_napi_cnt[cpu];
+ min_cpu = cpu;
+ }
+ }
+ }
+
+ if (min_cpu < 0)
+ return min_cpu;
+
+ octeon3_cpu_napi_cnt[min_cpu]++;
+
+ return min_cpu;
+}
+
+/* octeon3_napi_alloc - Allocate a napi.
+ * @cxt: Receive context the napi will be added to.
+ * @idx: Napi index within the receive context.
+ * @cpu: Cpu to bind the napi to:
+ * < 0: use any cpu.
+ * >= 0: use requested cpu.
+ *
+ * Returns pointer to napi wrapper.
+ * Returns NULL on error.
+ */
+static struct octeon3_napi_wrapper *octeon3_napi_alloc(struct octeon3_rx *cxt,
+ int idx, int cpu)
+{
+ struct octeon3_ethernet *priv = cxt->parent;
+ struct octeon3_ethernet_node *oen;
+ int i, node = priv->node;
+ unsigned long flags;
+
+ oen = octeon3_eth_node + node;
+ spin_lock_irqsave(&oen->napi_alloc_lock, flags);
+
+ /* Find a free napi wrapper */
+ for (i = 0; i < MAX_NAPIS_PER_NODE; i++) {
+ if (napi_wrapper[node][i].available) {
+ /* Allocate a cpu to use */
+ cpu = octeon3_napi_alloc_cpu(node, cpu);
+ if (cpu < 0)
+ break;
+
+ napi_wrapper[node][i].available = 0;
+ napi_wrapper[node][i].idx = idx;
+ napi_wrapper[node][i].cpu = cpu;
+ napi_wrapper[node][i].cxt = cxt;
+ spin_unlock_irqrestore(&oen->napi_alloc_lock, flags);
+ return &napi_wrapper[node][i];
+ }
+ }
+
+ spin_unlock_irqrestore(&oen->napi_alloc_lock, flags);
+ return NULL;
+}
+
+/* octeon_cpu_napi_sched - Schedule a napi for execution. The napi will start
+ * executing on the cpu calling this function.
+ * @info: Pointer to the napi to schedule for execution.
+ */
+static void octeon_cpu_napi_sched(void *info)
+{
+ struct napi_struct *napi = info;
+
+ napi_schedule(napi);
+}
+
+/* octeon3_rm_napi_from_cxt - Remove a napi from a receive context.
+ * @node: Node napi belongs to.
+ * @napiw: Pointer to napi to remove.
+ *
+ * Returns 0 if successful.
+ * Returns <0 for error codes.
+ */
+static int octeon3_rm_napi_from_cxt(int node,
+ struct octeon3_napi_wrapper *napiw)
+{
+ struct octeon3_ethernet_node *oen;
+ struct octeon3_rx *cxt;
+ unsigned long flags;
+ int idx;
+
+ oen = octeon3_eth_node + node;
+ cxt = napiw->cxt;
+ idx = napiw->idx;
+
+ /* Free the napi block */
+ spin_lock_irqsave(&oen->napi_alloc_lock, flags);
+ octeon3_cpu_napi_cnt[napiw->cpu]--;
+ napiw->available = 1;
+ napiw->idx = -1;
+ napiw->cpu = -1;
+ napiw->cxt = NULL;
+ spin_unlock_irqrestore(&oen->napi_alloc_lock, flags);
+
+ /* Free the napi idx */
+ spin_lock_irqsave(&cxt->napi_idx_lock, flags);
+ bitmap_clear(cxt->napi_idx_bitmap, idx, 1);
+ spin_unlock_irqrestore(&cxt->napi_idx_lock, flags);
+
+ return 0;
+}
+
+/* octeon3_add_napi_to_cxt - Add a napi to a receive context.
+ * @cxt: Pointer to receive context.
+ *
+ * Returns 0 if successful.
+ * Returns <0 for error codes.
+ */
+static int octeon3_add_napi_to_cxt(struct octeon3_rx *cxt)
+{
+ struct octeon3_ethernet *priv = cxt->parent;
+ struct octeon3_napi_wrapper *napiw;
+ unsigned long flags;
+ int idx, rc;
+
+ /* Get a free napi idx */
+ spin_lock_irqsave(&cxt->napi_idx_lock, flags);
+ idx = find_first_zero_bit(cxt->napi_idx_bitmap, MAX_CORES);
+ if (unlikely(idx >= MAX_CORES)) {
+ spin_unlock_irqrestore(&cxt->napi_idx_lock, flags);
+ return -ENOMEM;
+ }
+ bitmap_set(cxt->napi_idx_bitmap, idx, 1);
+ spin_unlock_irqrestore(&cxt->napi_idx_lock, flags);
+
+ /* Get a free napi block */
+ napiw = octeon3_napi_alloc(cxt, idx, -1);
+ if (unlikely(!napiw)) {
+ spin_lock_irqsave(&cxt->napi_idx_lock, flags);
+ bitmap_clear(cxt->napi_idx_bitmap, idx, 1);
+ spin_unlock_irqrestore(&cxt->napi_idx_lock, flags);
+ return -ENOMEM;
+ }
+
+ rc = smp_call_function_single(napiw->cpu, octeon_cpu_napi_sched,
+ &napiw->napi, 0);
+ if (unlikely(rc))
+ octeon3_rm_napi_from_cxt(priv->node, napiw);
+
+ return rc;
+}
+
+/* Receive one packet.
+ * returns the number of RX buffers consumed.
+ */
+static int octeon3_eth_rx_one(struct octeon3_rx *rx, bool is_async,
+ bool req_next)
+{
+ struct octeon3_ethernet *priv = rx->parent;
+ int len_remaining, ret, segments;
+ union buf_ptr packet_ptr;
+ unsigned int packet_len;
+ struct sk_buff *skb;
+ struct wqe *work;
+ struct wr_ret r;
+ void **buf;
+ u64 gaura;
+ u8 *data;
+
+ if (is_async)
+ r = octeon3_core_get_response_async();
+ else
+ r = octeon3_core_get_work_sync(rx->rx_grp);
+ work = r.work;
+ if (!work)
+ return 0;
+
+ /* Request the next work so it'll be ready when we need it */
+ if (is_async && req_next)
+ octeon3_core_get_work_async(rx->rx_grp);
+
+ skb = octeon3_eth_work_to_skb(work);
+
+ /* Save the aura and node this skb came from to allow the pko to free
+ * the skb back to the correct aura. A magic number is also added to
+ * later verify the skb came from the fpa.
+ *
+ * 63 12 11 10 9 0
+ * ---------------------------------------------------------------------
+ * | magic | node | aura |
+ * ---------------------------------------------------------------------
+ */
+ buf = (void **)PTR_ALIGN(skb->head, 128);
+ gaura = SKB_AURA_MAGIC | work->word0.aura;
+ buf[SKB_AURA_OFFSET] = (void *)gaura;
+
+ segments = work->word0.bufs;
+ ret = segments;
+ packet_ptr = work->packet_ptr;
+ if (unlikely(work->word2.err_level <= PKI_ERRLEV_LA &&
+ work->word2.err_code != PKI_OPCODE_NONE)) {
+ atomic64_inc(&priv->rx_errors);
+ switch (work->word2.err_code) {
+ case PKI_OPCODE_JABBER:
+ atomic64_inc(&priv->rx_length_errors);
+ break;
+ case PKI_OPCODE_FCS:
+ atomic64_inc(&priv->rx_crc_errors);
+ break;
+ }
+ data = phys_to_virt(packet_ptr.addr);
+ for (;;) {
+ dev_kfree_skb_any(skb);
+ segments--;
+ if (segments <= 0)
+ break;
+ packet_ptr.u64 = *(u64 *)(data - 8);
+#ifndef __LITTLE_ENDIAN
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X)) {
+ /* PKI_BUFLINK_S's are endian-swapped */
+ packet_ptr.u64 = swab64(packet_ptr.u64);
+ }
+#endif
+ data = phys_to_virt(packet_ptr.addr);
+ skb = octeon3_eth_work_to_skb((void *)round_down((unsigned long)data, 128ull));
+ }
+ goto out;
+ }
+
+ packet_len = work->word1.len;
+ data = phys_to_virt(packet_ptr.addr);
+ skb->data = data;
+ skb->len = packet_len;
+ len_remaining = packet_len;
+ if (segments == 1) {
+ /* Strip the ethernet fcs */
+ skb->len -= 4;
+ skb_set_tail_pointer(skb, skb->len);
+ } else {
+ bool first_frag = true;
+ struct sk_buff *current_skb = skb;
+ struct sk_buff *next_skb = NULL;
+ unsigned int segment_size;
+
+ skb_frag_list_init(skb);
+ for (;;) {
+ segment_size = (segments == 1) ?
+ len_remaining : packet_ptr.size;
+ len_remaining -= segment_size;
+ if (!first_frag) {
+ current_skb->len = segment_size;
+ skb->data_len += segment_size;
+ skb->truesize += current_skb->truesize;
+ }
+ skb_set_tail_pointer(current_skb, segment_size);
+ segments--;
+ if (segments == 0)
+ break;
+ packet_ptr.u64 = *(u64 *)(data - 8);
+#ifndef __LITTLE_ENDIAN
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X)) {
+ /* PKI_BUFLINK_S's are endian-swapped */
+ packet_ptr.u64 = swab64(packet_ptr.u64);
+ }
+#endif
+ data = phys_to_virt(packet_ptr.addr);
+ next_skb = octeon3_eth_work_to_skb((void *)round_down((unsigned long)data, 128ull));
+ if (first_frag) {
+ next_skb->next =
+ skb_shinfo(current_skb)->frag_list;
+ skb_shinfo(current_skb)->frag_list = next_skb;
+ } else {
+ current_skb->next = next_skb;
+ next_skb->next = NULL;
+ }
+ current_skb = next_skb;
+ first_frag = false;
+ current_skb->data = data;
+ }
+
+ /* Strip the ethernet fcs */
+ pskb_trim(skb, skb->len - 4);
+ }
+
+ if (likely(priv->netdev->flags & IFF_UP)) {
+ skb_checksum_none_assert(skb);
+ if (unlikely(priv->rx_timestamp_hw)) {
+ /* The first 8 bytes are the timestamp */
+ u64 hwts = *(u64 *)skb->data;
+ u64 ns;
+ struct skb_shared_hwtstamps *shts;
+
+ ns = timecounter_cyc2time(&priv->tc, hwts);
+ shts = skb_hwtstamps(skb);
+ memset(shts, 0, sizeof(*shts));
+ shts->hwtstamp = ns_to_ktime(ns);
+ __skb_pull(skb, 8);
+ }
+
+ skb->protocol = eth_type_trans(skb, priv->netdev);
+ skb->dev = priv->netdev;
+ if (priv->netdev->features & NETIF_F_RXCSUM) {
+ if ((work->word2.lc_hdr_type == PKI_LTYPE_IP4 ||
+ work->word2.lc_hdr_type == PKI_LTYPE_IP6) &&
+ (work->word2.lf_hdr_type == PKI_LTYPE_TCP ||
+ work->word2.lf_hdr_type == PKI_LTYPE_UDP ||
+ work->word2.lf_hdr_type == PKI_LTYPE_SCTP))
+ if (work->word2.err_code == 0)
+ skb->ip_summed = CHECKSUM_UNNECESSARY;
+ }
+
+ netif_receive_skb(skb);
+ } else {
+ /* Drop any packet received for a device that isn't up */
+ atomic64_inc(&priv->rx_dropped);
+ dev_kfree_skb_any(skb);
+ }
+out:
+ return ret;
+}
+
+static int octeon3_eth_napi(struct napi_struct *napi, int budget)
+{
+ int idx, napis_inuse, n = 0, n_bufs = 0, rx_count = 0;
+ struct octeon3_napi_wrapper *napiw;
+ struct octeon3_ethernet *priv;
+ u64 aq_cnt, old_scratch;
+ struct octeon3_rx *cxt;
+
+ napiw = container_of(napi, struct octeon3_napi_wrapper, napi);
+ cxt = napiw->cxt;
+ priv = cxt->parent;
+
+ /* Get the amount of work pending */
+ aq_cnt = oct_csr_read(SSO_GRP_AQ_CNT(priv->node, cxt->rx_grp));
+ aq_cnt &= GENMASK_ULL(32, 0);
+ /* Allow the last thread to add/remove threads if the work
+ * incremented/decremented by more than what the current number
+ * of threads can support.
+ */
+ idx = find_last_bit(cxt->napi_idx_bitmap, MAX_CORES);
+ napis_inuse = bitmap_weight(cxt->napi_idx_bitmap, MAX_CORES);
+
+ if (napiw->idx == idx) {
+ if (aq_cnt > napis_inuse * 128) {
+ octeon3_add_napi_to_cxt(cxt);
+ } else if (napiw->idx > 0 && aq_cnt < (napis_inuse - 1) * 128) {
+ napi_complete(napi);
+ octeon3_rm_napi_from_cxt(priv->node, napiw);
+ return 0;
+ }
+ }
+
+ if (likely(USE_ASYNC_IOBDMA)) {
+ /* Save scratch in case userspace is using it */
+ CVMX_SYNCIOBDMA;
+ old_scratch = scratch_read64(SCR_SCRATCH);
+
+ octeon3_core_get_work_async(cxt->rx_grp);
+ }
+
+ while (rx_count < budget) {
+ n = 0;
+
+ if (likely(USE_ASYNC_IOBDMA)) {
+ bool req_next = rx_count < (budget - 1) ? true : false;
+
+ n = octeon3_eth_rx_one(cxt, true, req_next);
+ } else {
+ n = octeon3_eth_rx_one(cxt, false, false);
+ }
+
+ if (n == 0)
+ break;
+
+ n_bufs += n;
+ rx_count++;
+ }
+
+ /* Wake up worker threads */
+ n_bufs = atomic64_add_return(n_bufs, &priv->buffers_needed);
+ if (n_bufs >= 32) {
+ struct octeon3_ethernet_node *oen;
+
+ oen = octeon3_eth_node + priv->node;
+ atomic_set(&oen->workers[0].kick, 1);
+ wake_up(&oen->workers[0].queue);
+ }
+
+ /* Stop the thread when no work is pending */
+ if (rx_count < budget) {
+ napi_complete(napi);
+
+ if (napiw->idx > 0)
+ octeon3_rm_napi_from_cxt(priv->node, napiw);
+ else
+ octeon3_sso_irq_set(cxt->parent->node, cxt->rx_grp,
+ true);
+ }
+
+ if (likely(USE_ASYNC_IOBDMA)) {
+ /* Restore the scratch area */
+ scratch_write64(SCR_SCRATCH, old_scratch);
+ }
+
+ return rx_count;
+}
+
+/* octeon3_napi_init_node - Initialize the node napis.
+ * @node: Node napis belong to.
+ * @netdev: Default network device used to initialize the napis.
+ *
+ * Returns 0 if successful.
+ * Returns <0 for error codes.
+ */
+static int octeon3_napi_init_node(int node, struct net_device *netdev)
+{
+ struct octeon3_ethernet_node *oen;
+ unsigned long flags;
+ int i;
+
+ oen = octeon3_eth_node + node;
+ spin_lock_irqsave(&oen->napi_alloc_lock, flags);
+
+ if (oen->napi_init_done)
+ goto done;
+
+ for (i = 0; i < MAX_NAPIS_PER_NODE; i++) {
+ netif_napi_add(netdev, &napi_wrapper[node][i].napi,
+ octeon3_eth_napi, 32);
+ napi_enable(&napi_wrapper[node][i].napi);
+ napi_wrapper[node][i].available = 1;
+ napi_wrapper[node][i].idx = -1;
+ napi_wrapper[node][i].cpu = -1;
+ napi_wrapper[node][i].cxt = NULL;
+ }
+
+ oen->napi_init_done = true;
+done:
+ spin_unlock_irqrestore(&oen->napi_alloc_lock, flags);
+ return 0;
+}
+
+#undef BROKEN_SIMULATOR_CSUM
+
+static void ethtool_get_drvinfo(struct net_device *netdev,
+ struct ethtool_drvinfo *info)
+{
+ strcpy(info->driver, "octeon3-ethernet");
+ strcpy(info->version, "1.0");
+ strcpy(info->bus_info, "Builtin");
+}
+
+static int ethtool_get_ts_info(struct net_device *ndev,
+ struct ethtool_ts_info *info)
+{
+ struct octeon3_ethernet *priv = netdev_priv(ndev);
+
+ info->so_timestamping = SOF_TIMESTAMPING_TX_HARDWARE |
+ SOF_TIMESTAMPING_RX_HARDWARE | SOF_TIMESTAMPING_RAW_HARDWARE;
+
+ if (priv->ptp_clock)
+ info->phc_index = ptp_clock_index(priv->ptp_clock);
+ else
+ info->phc_index = -1;
+
+ info->tx_types = (1 << HWTSTAMP_TX_OFF) | (1 << HWTSTAMP_TX_ON);
+
+ info->rx_filters = (1 << HWTSTAMP_FILTER_NONE) |
+ (1 << HWTSTAMP_FILTER_ALL);
+
+ return 0;
+}
+
+static const struct ethtool_ops octeon3_ethtool_ops = {
+ .get_drvinfo = ethtool_get_drvinfo,
+ .get_link_ksettings = bgx_port_ethtool_get_link_ksettings,
+ .set_settings = bgx_port_ethtool_set_settings,
+ .nway_reset = bgx_port_ethtool_nway_reset,
+ .get_link = ethtool_op_get_link,
+ .get_ts_info = ethtool_get_ts_info,
+};
+
+static int octeon3_eth_ndo_change_mtu(struct net_device *netdev, int new_mtu)
+{
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X)) {
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ int fifo_size, max_mtu = 1500;
+
+ /* On 78XX-Pass1 the mtu must be limited. The PKO may
+ * to lock up when calculating the L4 checksum for
+ * large packets. How large the packets can be depends
+ * on the amount of pko fifo assigned to the port.
+ *
+ * FIFO size Max frame size
+ * 2.5 KB 1920
+ * 5.0 KB 4480
+ * 10.0 KB 9600
+ *
+ * The maximum mtu is set to the largest frame size minus the
+ * l2 header.
+ */
+ fifo_size = octeon3_pko_get_fifo_size(priv->node,
+ priv->interface,
+ priv->index,
+ priv->mac_type);
+
+ switch (fifo_size) {
+ case 2560:
+ max_mtu = 1920 - ETH_HLEN - ETH_FCS_LEN -
+ (2 * VLAN_HLEN);
+ break;
+
+ case 5120:
+ max_mtu = 4480 - ETH_HLEN - ETH_FCS_LEN -
+ (2 * VLAN_HLEN);
+ break;
+
+ case 10240:
+ max_mtu = 9600 - ETH_HLEN - ETH_FCS_LEN -
+ (2 * VLAN_HLEN);
+ break;
+
+ default:
+ break;
+ }
+ if (new_mtu > max_mtu) {
+ netdev_warn(netdev, "Maximum MTU supported is %d",
+ max_mtu);
+ return -EINVAL;
+ }
+ }
+ return bgx_port_change_mtu(netdev, new_mtu);
+}
+
+static int octeon3_eth_common_ndo_init(struct net_device *netdev,
+ int extra_skip)
+{
+ int aura, base_rx_grp[MAX_RX_CONTEXTS], dq, i, pki_chan, r;
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ struct octeon3_ethernet_node *oen = octeon3_eth_node + priv->node;
+
+ netif_carrier_off(netdev);
+
+ netdev->features |=
+#ifndef BROKEN_SIMULATOR_CSUM
+ NETIF_F_IP_CSUM |
+ NETIF_F_IPV6_CSUM |
+#endif
+ NETIF_F_SG |
+ NETIF_F_FRAGLIST |
+ NETIF_F_RXCSUM |
+ NETIF_F_LLTX;
+
+ if (!OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X))
+ netdev->features |= NETIF_F_SCTP_CRC;
+
+ netdev->features |= NETIF_F_TSO | NETIF_F_TSO6;
+
+ /* Set user changeable settings */
+ netdev->hw_features = netdev->features;
+
+ priv->rx_buf_count = num_packet_buffers;
+
+ pki_chan = get_pki_chan(priv->node, priv->interface, priv->index);
+
+ dq = octeon3_pko_interface_init(priv->node, priv->interface,
+ priv->index, priv->mac_type, pki_chan);
+ if (dq < 0) {
+ dev_err(netdev->dev.parent, "Failed to initialize pko\n");
+ return -ENODEV;
+ }
+
+ r = octeon3_pko_activate_dq(priv->node, dq, 1);
+ if (r < 0) {
+ dev_err(netdev->dev.parent, "Failed to activate dq\n");
+ return -ENODEV;
+ }
+
+ priv->pko_queue = dq;
+ octeon_fpa3_aura_init(priv->node, oen->pki_packet_pool, -1, &aura,
+ num_packet_buffers, num_packet_buffers * 2);
+ priv->pki_aura = aura;
+ aura2bufs_needed[priv->node][priv->pki_aura] = &priv->buffers_needed;
+
+ r = octeon3_sso_alloc_groups(priv->node, base_rx_grp, rx_contexts, -1);
+ if (r) {
+ dev_err(netdev->dev.parent, "Failed to allocated SSO group\n");
+ return -ENODEV;
+ }
+ for (i = 0; i < rx_contexts; i++) {
+ priv->rx_cxt[i].rx_grp = base_rx_grp[i];
+ priv->rx_cxt[i].parent = priv;
+
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X))
+ octeon3_sso_pass1_limit(priv->node,
+ priv->rx_cxt[i].rx_grp);
+ }
+ priv->num_rx_cxt = rx_contexts;
+
+ priv->tx_complete_grp = oen->tx_complete_grp;
+ dev_info(netdev->dev.parent,
+ "rx sso grp:%d..%d aura:%d pknd:%d pko_queue:%d\n",
+ *base_rx_grp, *(base_rx_grp + priv->num_rx_cxt - 1),
+ priv->pki_aura, priv->pknd, priv->pko_queue);
+
+ octeon3_pki_port_init(priv->node, priv->pki_aura, *base_rx_grp,
+ extra_skip, (packet_buffer_size - 128),
+ priv->pknd, priv->num_rx_cxt);
+
+ priv->last_packets = 0;
+ priv->last_octets = 0;
+ priv->last_dropped = 0;
+
+ octeon3_napi_init_node(priv->node, netdev);
+
+ /* Register ethtool methods */
+ netdev->ethtool_ops = &octeon3_ethtool_ops;
+
+ return 0;
+}
+
+static int octeon3_eth_bgx_ndo_init(struct net_device *netdev)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ const u8 *mac;
+ int r;
+
+ priv->pknd = bgx_port_get_pknd(priv->node, priv->interface,
+ priv->index);
+ octeon3_eth_common_ndo_init(netdev, 0);
+
+ /* Padding and FCS are done in BGX */
+ r = octeon3_pko_set_mac_options(priv->node, priv->interface,
+ priv->index, priv->mac_type, false,
+ false, 0);
+ if (r)
+ return r;
+
+ mac = bgx_port_get_mac(netdev);
+ if (mac && is_valid_ether_addr(mac)) {
+ memcpy(netdev->dev_addr, mac, ETH_ALEN);
+ netdev->addr_assign_type &= ~NET_ADDR_RANDOM;
+ } else {
+ eth_hw_addr_random(netdev);
+ }
+
+ bgx_port_set_rx_filtering(netdev);
+ octeon3_eth_ndo_change_mtu(netdev, netdev->mtu);
+
+ return 0;
+}
+
+static void octeon3_eth_ndo_uninit(struct net_device *netdev)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ int grp[MAX_RX_CONTEXTS], i;
+
+ /* Shutdwon pki for this interface */
+ octeon3_pki_port_shutdown(priv->node, priv->pknd);
+ octeon_fpa3_release_aura(priv->node, priv->pki_aura);
+ aura2bufs_needed[priv->node][priv->pki_aura] = NULL;
+
+ /* Shutdown pko for this interface */
+ octeon3_pko_interface_uninit(priv->node, &priv->pko_queue, 1);
+
+ /* Free the receive contexts sso groups */
+ for (i = 0; i < rx_contexts; i++)
+ grp[i] = priv->rx_cxt[i].rx_grp;
+ octeon3_sso_free_groups(priv->node, grp, rx_contexts);
+}
+
+static irqreturn_t octeon3_eth_rx_handler(int irq, void *info)
+{
+ struct octeon3_rx *rx = info;
+
+ /* Disarm the irq. */
+ octeon3_sso_irq_set(rx->parent->node, rx->rx_grp, false);
+
+ napi_schedule(&rx->napiw->napi);
+ return IRQ_HANDLED;
+}
+
+static int octeon3_eth_common_ndo_open(struct net_device *netdev)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ struct octeon3_rx *rx;
+ int i, idx, r;
+
+ for (i = 0; i < priv->num_rx_cxt; i++) {
+ unsigned int sso_intsn;
+ int cpu;
+
+ rx = priv->rx_cxt + i;
+ sso_intsn = SSO_INTSN_EXE << 12 | rx->rx_grp;
+
+ spin_lock_init(&rx->napi_idx_lock);
+
+ rx->rx_irq = irq_create_mapping(NULL, sso_intsn);
+ if (!rx->rx_irq) {
+ netdev_err(netdev, "ERROR: Couldn't map hwirq: %x\n",
+ sso_intsn);
+ r = -EINVAL;
+ goto err1;
+ }
+ r = request_irq(rx->rx_irq, octeon3_eth_rx_handler,
+ IRQ_TYPE_EDGE_RISING, netdev_name(netdev), rx);
+ if (r) {
+ netdev_err(netdev, "ERROR: Couldn't request irq: %d\n",
+ rx->rx_irq);
+ r = -ENOMEM;
+ goto err2;
+ }
+
+ octeon3_eth_gen_affinity(priv->node, &rx->rx_affinity_hint);
+ irq_set_affinity_hint(rx->rx_irq, &rx->rx_affinity_hint);
+
+ /* Allocate a napi index for this receive context */
+ bitmap_zero(priv->rx_cxt[i].napi_idx_bitmap, MAX_CORES);
+ idx = find_first_zero_bit(priv->rx_cxt[i].napi_idx_bitmap,
+ MAX_CORES);
+ if (idx >= MAX_CORES) {
+ netdev_err(netdev, "ERROR: Couldn't get napi index\n");
+ r = -ENOMEM;
+ goto err3;
+ }
+ bitmap_set(priv->rx_cxt[i].napi_idx_bitmap, idx, 1);
+ cpu = cpumask_first(&rx->rx_affinity_hint);
+
+ priv->rx_cxt[i].napiw = octeon3_napi_alloc(&priv->rx_cxt[i],
+ idx, cpu);
+ if (!priv->rx_cxt[i].napiw) {
+ r = -ENOMEM;
+ goto err4;
+ }
+
+ /* Arm the irq. */
+ octeon3_sso_irq_set(priv->node, rx->rx_grp, true);
+ }
+ octeon3_eth_replenish_rx(priv, priv->rx_buf_count);
+
+ return 0;
+
+err4:
+ bitmap_clear(priv->rx_cxt[i].napi_idx_bitmap, idx, 1);
+err3:
+ irq_set_affinity_hint(rx->rx_irq, NULL);
+ free_irq(rx->rx_irq, rx);
+err2:
+ irq_dispose_mapping(rx->rx_irq);
+err1:
+ for (i--; i >= 0; i--) {
+ rx = priv->rx_cxt + i;
+ irq_dispose_mapping(rx->rx_irq);
+ free_irq(rx->rx_irq, rx);
+ octeon3_rm_napi_from_cxt(priv->node, priv->rx_cxt[i].napiw);
+ priv->rx_cxt[i].napiw = NULL;
+ }
+
+ return r;
+}
+
+static int octeon3_eth_bgx_ndo_open(struct net_device *netdev)
+{
+ int rc;
+
+ rc = octeon3_eth_common_ndo_open(netdev);
+ if (rc == 0)
+ rc = bgx_port_enable(netdev);
+
+ return rc;
+}
+
+static int octeon3_eth_common_ndo_stop(struct net_device *netdev)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ struct octeon3_rx *rx;
+ struct sk_buff *skb;
+ void **w;
+ int i;
+
+ /* Allow enough time for ingress in transit packets to be drained */
+ msleep(20);
+
+ /* Wait until sso has no more work for this interface */
+ for (i = 0; i < priv->num_rx_cxt; i++) {
+ rx = priv->rx_cxt + i;
+ while (oct_csr_read(SSO_GRP_AQ_CNT(priv->node, rx->rx_grp)))
+ msleep(20);
+ }
+
+ /* Free the irq and napi context for each rx context */
+ for (i = 0; i < priv->num_rx_cxt; i++) {
+ rx = priv->rx_cxt + i;
+ octeon3_sso_irq_set(priv->node, rx->rx_grp, false);
+ irq_set_affinity_hint(rx->rx_irq, NULL);
+ free_irq(rx->rx_irq, rx);
+ irq_dispose_mapping(rx->rx_irq);
+ rx->rx_irq = 0;
+
+ octeon3_rm_napi_from_cxt(priv->node, rx->napiw);
+ rx->napiw = NULL;
+ WARN_ON(!bitmap_empty(rx->napi_idx_bitmap, MAX_CORES));
+ }
+
+ /* Free the packet buffers */
+ for (;;) {
+ w = octeon_fpa3_alloc(priv->node, priv->pki_aura);
+ if (!w)
+ break;
+ skb = w[0];
+ dev_kfree_skb(skb);
+ }
+
+ return 0;
+}
+
+static int octeon3_eth_bgx_ndo_stop(struct net_device *netdev)
+{
+ int r;
+
+ r = bgx_port_disable(netdev);
+ if (r)
+ return r;
+
+ return octeon3_eth_common_ndo_stop(netdev);
+}
+
+static inline u64 build_pko_send_hdr_desc(struct sk_buff *skb, int gaura)
+{
+ u64 checksum_alg, send_hdr = 0;
+ u8 l4_hdr = 0;
+
+ /* See PKO_SEND_HDR_S in the HRM for the send header descriptor
+ * format.
+ */
+#ifdef __LITTLE_ENDIAN
+ send_hdr |= BIT(43);
+#endif
+
+ if (!OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X)) {
+ /* Don't allocate to L2 */
+ send_hdr |= BIT(42);
+ }
+
+ /* Don't automatically free to FPA */
+ send_hdr |= BIT(40);
+
+ send_hdr |= skb->len;
+ send_hdr |= (u64)gaura << 48;
+
+ if (skb->ip_summed != CHECKSUM_NONE &&
+ skb->ip_summed != CHECKSUM_UNNECESSARY) {
+#ifndef BROKEN_SIMULATOR_CSUM
+ switch (skb->protocol) {
+ case htons(ETH_P_IP):
+ send_hdr |= ETH_HLEN << 16;
+ send_hdr |= BIT(45);
+ l4_hdr = ip_hdr(skb)->protocol;
+ send_hdr |= (ETH_HLEN + (4 * ip_hdr(skb)->ihl)) << 24;
+ break;
+
+ case htons(ETH_P_IPV6):
+ l4_hdr = ipv6_hdr(skb)->nexthdr;
+ send_hdr |= ETH_HLEN << 16;
+ break;
+
+ default:
+ break;
+ }
+#endif
+
+ checksum_alg = 1; /* UDP == 1 */
+ switch (l4_hdr) {
+ case IPPROTO_SCTP:
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX_PASS1_X))
+ break;
+ checksum_alg++; /* SCTP == 3 */
+ /* Fall through */
+ case IPPROTO_TCP: /* TCP == 2 */
+ checksum_alg++;
+ /* Fall through */
+ case IPPROTO_UDP:
+ if (skb_transport_header_was_set(skb)) {
+ int l4ptr = skb_transport_header(skb) -
+ skb->data;
+ send_hdr &= ~GENMASK_ULL(31, 24);
+ send_hdr |= l4ptr << 24;
+ send_hdr |= checksum_alg << 46;
+ }
+ break;
+
+ default:
+ break;
+ }
+ }
+
+ return send_hdr;
+}
+
+static inline u64 build_pko_send_ext_desc(struct sk_buff *skb)
+{
+ u64 send_ext = 0;
+
+ /* See PKO_SEND_EXT_S in the HRM for the send extended descriptor
+ * format.
+ */
+ skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
+ send_ext |= (u64)PKO_SENDSUBDC_EXT << 44;
+ send_ext |= 1ull << 40;
+ send_ext |= BIT(39);
+ send_ext |= ETH_HLEN << 16;
+
+ return send_ext;
+}
+
+static inline u64 build_pko_send_tso(struct sk_buff *skb, uint mtu)
+{
+ u64 send_tso = 0;
+
+ /* See PKO_SEND_TSO_S in the HRM for the send tso descriptor format */
+ send_tso |= 12ull << 56;
+ send_tso |= (u64)PKO_SENDSUBDC_TSO << 44;
+ send_tso |= (skb_transport_offset(skb) + tcp_hdrlen(skb)) << 24;
+ send_tso |= (mtu + ETH_HLEN) << 8;
+
+ return send_tso;
+}
+
+static inline u64 build_pko_send_mem_sub(u64 addr)
+{
+ u64 send_mem = 0;
+
+ /* See PKO_SEND_MEM_S in the HRM for the send mem descriptor format */
+ send_mem |= (u64)PKO_SENDSUBDC_MEM << 44;
+ send_mem |= (u64)MEMDSZ_B64 << 60;
+ send_mem |= (u64)MEMALG_SUB << 56;
+ send_mem |= 1ull << 48;
+ send_mem |= addr;
+
+ return send_mem;
+}
+
+static inline u64 build_pko_send_mem_ts(u64 addr)
+{
+ u64 send_mem = 0;
+
+ /* See PKO_SEND_MEM_S in the HRM for the send mem descriptor format */
+ send_mem |= 1ull << 62;
+ send_mem |= (u64)PKO_SENDSUBDC_MEM << 44;
+ send_mem |= (u64)MEMDSZ_B64 << 60;
+ send_mem |= (u64)MEMALG_SETTSTMP << 56;
+ send_mem |= addr;
+
+ return send_mem;
+}
+
+static inline u64 build_pko_send_free(u64 addr)
+{
+ u64 send_free = 0;
+
+ /* See PKO_SEND_FREE_S in the HRM for the send free descriptor format */
+ send_free |= (u64)PKO_SENDSUBDC_FREE << 44;
+ send_free |= addr;
+
+ return send_free;
+}
+
+static inline u64 build_pko_send_work(int grp, u64 addr)
+{
+ u64 send_work = 0;
+
+ /* See PKO_SEND_WORK_S in the HRM for the send work descriptor format */
+ send_work |= (u64)PKO_SENDSUBDC_WORK << 44;
+ send_work |= (u64)grp << 52;
+ send_work |= 2ull << 50;
+ send_work |= addr;
+
+ return send_work;
+}
+
+static int octeon3_eth_ndo_start_xmit(struct sk_buff *skb,
+ struct net_device *netdev)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ u64 aq_cnt = 0, *dma_addr, head_len, lmtdma_data;
+ u64 pko_send_desc, scr_off = LMTDMA_SCR_OFFSET;
+ int frag_count, gaura = 0, grp, i;
+ struct octeon3_ethernet_node *oen;
+ struct sk_buff *skb_tmp;
+ unsigned int mss;
+ long backlog;
+ void **work;
+
+ frag_count = 0;
+ if (skb_has_frag_list(skb))
+ skb_walk_frags(skb, skb_tmp)
+ frag_count++;
+
+ /* Drop the packet if pko or sso are not keeping up */
+ oen = octeon3_eth_node + priv->node;
+ aq_cnt = oct_csr_read(SSO_GRP_AQ_CNT(oen->node, oen->tx_complete_grp));
+ aq_cnt &= GENMASK_ULL(32, 0);
+ backlog = atomic64_inc_return(&priv->tx_backlog);
+ if (unlikely(backlog > MAX_TX_QUEUE_DEPTH || aq_cnt > 100000)) {
+ if (use_tx_queues) {
+ netif_stop_queue(netdev);
+ } else {
+ atomic64_dec(&priv->tx_backlog);
+ goto skip_xmit;
+ }
+ }
+
+ /* We have space for 11 segment pointers, If there will be
+ * more than that, we must linearize. The count is: 1 (base
+ * SKB) + frag_count + nr_frags.
+ */
+ if (unlikely(skb_shinfo(skb)->nr_frags + frag_count > 10)) {
+ if (unlikely(__skb_linearize(skb)))
+ goto skip_xmit;
+ frag_count = 0;
+ }
+
+ work = (void **)skb->cb;
+ work[0] = netdev;
+ work[1] = NULL;
+
+ /* Adjust the port statistics. */
+ atomic64_inc(&priv->tx_packets);
+ atomic64_add(skb->len, &priv->tx_octets);
+
+ /* Make sure packet data writes are committed before
+ * submitting the command below
+ */
+ wmb();
+
+ /* Build the pko command */
+ pko_send_desc = build_pko_send_hdr_desc(skb, gaura);
+ preempt_disable();
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+
+ /* Request packet to be ptp timestamped */
+ if ((unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) &&
+ unlikely(priv->tx_timestamp_hw)) {
+ pko_send_desc = build_pko_send_ext_desc(skb);
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+ }
+
+ /* Add the tso descriptor if needed */
+ mss = skb_shinfo(skb)->gso_size;
+ if (unlikely(mss)) {
+ pko_send_desc = build_pko_send_tso(skb, netdev->mtu);
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+ }
+
+ /* Add a gather descriptor for each segment. See PKO_SEND_GATHER_S for
+ * the send gather descriptor format.
+ */
+ pko_send_desc = 0;
+ pko_send_desc |= (u64)PKO_SENDSUBDC_GATHER << 45;
+ head_len = skb_headlen(skb);
+ if (head_len > 0) {
+ pko_send_desc |= head_len << 48;
+ pko_send_desc |= virt_to_phys(skb->data);
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+ }
+ for (i = 1; i <= skb_shinfo(skb)->nr_frags; i++) {
+ struct skb_frag_struct *fs = skb_shinfo(skb)->frags + i - 1;
+
+ pko_send_desc &= ~(GENMASK_ULL(63, 48) | GENMASK_ULL(41, 0));
+ pko_send_desc |= (u64)fs->size << 48;
+ pko_send_desc |= virt_to_phys((u8 *)page_address(fs->page.p) +
+ fs->page_offset);
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+ }
+ skb_walk_frags(skb, skb_tmp) {
+ pko_send_desc &= ~(GENMASK_ULL(63, 48) | GENMASK_ULL(41, 0));
+ pko_send_desc |= (u64)skb_tmp->len << 48;
+ pko_send_desc |= virt_to_phys(skb_tmp->data);
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+ }
+
+ /* Subtract 1 from the tx_backlog. */
+ pko_send_desc = build_pko_send_mem_sub(virt_to_phys(&priv->tx_backlog));
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+
+ /* Write the ptp timestamp in the skb itself */
+ if ((unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP)) &&
+ unlikely(priv->tx_timestamp_hw)) {
+ pko_send_desc = build_pko_send_mem_ts(virt_to_phys(&work[1]));
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+ }
+
+ /* Send work when finished with the packet. */
+ grp = octeon3_eth_lgrp_to_ggrp(priv->node, priv->tx_complete_grp);
+ pko_send_desc = build_pko_send_work(grp, virt_to_phys(work));
+ scratch_write64(scr_off, pko_send_desc);
+ scr_off += sizeof(pko_send_desc);
+
+ /* See PKO_SEND_DMA_S in the HRM for the lmtdam data format */
+ lmtdma_data = 0;
+ lmtdma_data |= (u64)(LMTDMA_SCR_OFFSET >> 3) << 56;
+ if (wait_pko_response)
+ lmtdma_data |= 1ull << 48;
+ lmtdma_data |= 0x51ull << 40;
+ lmtdma_data |= (u64)priv->node << 36;
+ lmtdma_data |= priv->pko_queue << 16;
+
+ dma_addr = (u64 *)(LMTDMA_ORDERED_IO_ADDR | ((scr_off & 0x78) - 8));
+ *dma_addr = lmtdma_data;
+
+ preempt_enable();
+
+ if (wait_pko_response) {
+ u64 query_rtn;
+
+ CVMX_SYNCIOBDMA;
+
+ /* See PKO_QUERY_RTN_S in the HRM for the return format */
+ query_rtn = scratch_read64(LMTDMA_SCR_OFFSET);
+ query_rtn >>= 60;
+ if (unlikely(query_rtn != PKO_DQSTATUS_PASS)) {
+ netdev_err(netdev, "PKO enqueue failed %llx\n",
+ (unsigned long long)query_rtn);
+ dev_kfree_skb_any(skb);
+ }
+ }
+
+ return NETDEV_TX_OK;
+skip_xmit:
+ atomic64_inc(&priv->tx_dropped);
+ dev_kfree_skb_any(skb);
+ return NETDEV_TX_OK;
+}
+
+static void octeon3_eth_ndo_get_stats64(struct net_device *netdev,
+ struct rtnl_link_stats64 *s)
+{
+ u64 delta_dropped, delta_octets, delta_packets, dropped;
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ u64 octets, packets;
+
+ spin_lock(&priv->stat_lock);
+
+ octeon3_pki_get_stats(priv->node, priv->pknd, &packets, &octets,
+ &dropped);
+
+ delta_packets = (packets - priv->last_packets) & ((1ull << 48) - 1);
+ delta_octets = (octets - priv->last_octets) & ((1ull << 48) - 1);
+ delta_dropped = (dropped - priv->last_dropped) & ((1ull << 48) - 1);
+
+ priv->last_packets = packets;
+ priv->last_octets = octets;
+ priv->last_dropped = dropped;
+
+ spin_unlock(&priv->stat_lock);
+
+ atomic64_add(delta_packets, &priv->rx_packets);
+ atomic64_add(delta_octets, &priv->rx_octets);
+ atomic64_add(delta_dropped, &priv->rx_dropped);
+
+ s->rx_packets = atomic64_read(&priv->rx_packets);
+ s->rx_bytes = atomic64_read(&priv->rx_octets);
+ s->rx_dropped = atomic64_read(&priv->rx_dropped);
+ s->rx_errors = atomic64_read(&priv->rx_errors);
+ s->rx_length_errors = atomic64_read(&priv->rx_length_errors);
+ s->rx_crc_errors = atomic64_read(&priv->rx_crc_errors);
+
+ s->tx_packets = atomic64_read(&priv->tx_packets);
+ s->tx_bytes = atomic64_read(&priv->tx_octets);
+ s->tx_dropped = atomic64_read(&priv->tx_dropped);
+}
+
+static int octeon3_eth_set_mac_address(struct net_device *netdev, void *addr)
+{
+ int r = eth_mac_addr(netdev, addr);
+
+ if (r)
+ return r;
+
+ bgx_port_set_rx_filtering(netdev);
+
+ return 0;
+}
+
+static u64 octeon3_cyclecounter_read(const struct cyclecounter *cc)
+{
+ struct octeon3_ethernet *priv;
+ u64 count;
+
+ priv = container_of(cc, struct octeon3_ethernet, cc);
+ count = oct_csr_read(MIO_PTP_CLOCK_HI(priv->node));
+ return count;
+}
+
+static int octeon3_bgx_hwtstamp(struct net_device *netdev, int en)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ u64 data;
+
+ switch (bgx_port_get_mode(priv->node, priv->interface, priv->index)) {
+ case PORT_MODE_RGMII:
+ case PORT_MODE_SGMII:
+ data = oct_csr_read(BGX_GMP_GMI_RX_FRM_CTL(priv->node,
+ priv->interface,
+ priv->index));
+ if (en)
+ data |= BIT(12);
+ else
+ data &= ~BIT(12);
+ oct_csr_write(data, BGX_GMP_GMI_RX_FRM_CTL(priv->node,
+ priv->interface,
+ priv->index));
+ break;
+
+ case PORT_MODE_XAUI:
+ case PORT_MODE_RXAUI:
+ case PORT_MODE_10G_KR:
+ case PORT_MODE_XLAUI:
+ case PORT_MODE_40G_KR4:
+ case PORT_MODE_XFI:
+ data = oct_csr_read(BGX_SMU_RX_FRM_CTL(priv->node,
+ priv->interface,
+ priv->index));
+ if (en)
+ data |= BIT(12);
+ else
+ data &= ~BIT(12);
+ oct_csr_write(data, BGX_SMU_RX_FRM_CTL(priv->node,
+ priv->interface,
+ priv->index));
+ break;
+
+ default:
+ /* No timestamp support*/
+ return -EOPNOTSUPP;
+ }
+
+ return 0;
+}
+
+static int octeon3_pki_hwtstamp(struct net_device *netdev, int en)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ int skip = en ? 8 : 0;
+
+ octeon3_pki_set_ptp_skip(priv->node, priv->pknd, skip);
+
+ return 0;
+}
+
+static int octeon3_ioctl_hwtstamp(struct net_device *netdev, struct ifreq *rq,
+ int cmd)
+{
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ struct hwtstamp_config config;
+ u64 data;
+ int en;
+
+ /* The PTP block should be enabled */
+ data = oct_csr_read(MIO_PTP_CLOCK_CFG(priv->node));
+ if (!(data & BIT(0))) {
+ netdev_err(netdev, "Error: PTP clock not enabled\n");
+ return -EOPNOTSUPP;
+ }
+
+ if (copy_from_user(&config, rq->ifr_data, sizeof(config)))
+ return -EFAULT;
+
+ if (config.flags) /* reserved for future extensions */
+ return -EINVAL;
+
+ switch (config.tx_type) {
+ case HWTSTAMP_TX_OFF:
+ priv->tx_timestamp_hw = 0;
+ break;
+ case HWTSTAMP_TX_ON:
+ priv->tx_timestamp_hw = 1;
+ break;
+ default:
+ return -ERANGE;
+ }
+
+ switch (config.rx_filter) {
+ case HWTSTAMP_FILTER_NONE:
+ priv->rx_timestamp_hw = 0;
+ en = 0;
+ break;
+ case HWTSTAMP_FILTER_ALL:
+ case HWTSTAMP_FILTER_SOME:
+ case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
+ priv->rx_timestamp_hw = 1;
+ en = 1;
+ break;
+ default:
+ return -ERANGE;
+ }
+
+ octeon3_bgx_hwtstamp(netdev, en);
+ octeon3_pki_hwtstamp(netdev, en);
+
+ priv->cc.read = octeon3_cyclecounter_read;
+ priv->cc.mask = CYCLECOUNTER_MASK(64);
+ /* Ptp counter is always in nsec */
+ priv->cc.mult = 1;
+ priv->cc.shift = 0;
+ timecounter_init(&priv->tc, &priv->cc, ktime_to_ns(ktime_get_real()));
+
+ return 0;
+}
+
+static int octeon3_adjfreq(struct ptp_clock_info *ptp, s32 ppb)
+{
+ struct octeon3_ethernet *priv;
+ int neg_ppb = 0;
+ u64 comp, diff;
+
+ priv = container_of(ptp, struct octeon3_ethernet, ptp_info);
+
+ if (ppb < 0) {
+ ppb = -ppb;
+ neg_ppb = 1;
+ }
+
+ /* The part per billion (ppb) is a delta from the base frequency */
+ comp = (NSEC_PER_SEC << 32) / octeon_get_io_clock_rate();
+
+ diff = comp;
+ diff *= ppb;
+ diff = div_u64(diff, 1000000000ULL);
+
+ comp = neg_ppb ? comp - diff : comp + diff;
+
+ oct_csr_write(comp, MIO_PTP_CLOCK_COMP(priv->node));
+
+ return 0;
+}
+
+static int octeon3_adjtime(struct ptp_clock_info *ptp, s64 delta)
+{
+ struct octeon3_ethernet *priv;
+ unsigned long flags;
+ s64 now;
+
+ priv = container_of(ptp, struct octeon3_ethernet, ptp_info);
+
+ spin_lock_irqsave(&priv->ptp_lock, flags);
+ now = timecounter_read(&priv->tc);
+ now += delta;
+ timecounter_init(&priv->tc, &priv->cc, now);
+ spin_unlock_irqrestore(&priv->ptp_lock, flags);
+
+ return 0;
+}
+
+static int octeon3_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
+{
+ struct octeon3_ethernet *priv;
+ unsigned long flags;
+ u32 remainder;
+ u64 ns;
+
+ priv = container_of(ptp, struct octeon3_ethernet, ptp_info);
+
+ spin_lock_irqsave(&priv->ptp_lock, flags);
+ ns = timecounter_read(&priv->tc);
+ spin_unlock_irqrestore(&priv->ptp_lock, flags);
+ ts->tv_sec = div_u64_rem(ns, 1000000000ULL, &remainder);
+ ts->tv_nsec = remainder;
+
+ return 0;
+}
+
+static int octeon3_settime(struct ptp_clock_info *ptp,
+ const struct timespec *ts)
+{
+ struct octeon3_ethernet *priv;
+ unsigned long flags;
+ u64 ns;
+
+ priv = container_of(ptp, struct octeon3_ethernet, ptp_info);
+ ns = timespec_to_ns(ts);
+
+ spin_lock_irqsave(&priv->ptp_lock, flags);
+ timecounter_init(&priv->tc, &priv->cc, ns);
+ spin_unlock_irqrestore(&priv->ptp_lock, flags);
+
+ return 0;
+}
+
+static int octeon3_enable(struct ptp_clock_info *ptp,
+ struct ptp_clock_request *rq, int on)
+{
+ return -EOPNOTSUPP;
+}
+
+static int octeon3_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
+{
+ int rc;
+
+ switch (cmd) {
+ case SIOCSHWTSTAMP:
+ rc = octeon3_ioctl_hwtstamp(netdev, ifr, cmd);
+ break;
+
+ default:
+ rc = bgx_port_do_ioctl(netdev, ifr, cmd);
+ break;
+ }
+
+ return rc;
+}
+
+static const struct net_device_ops octeon3_eth_netdev_ops = {
+ .ndo_init = octeon3_eth_bgx_ndo_init,
+ .ndo_uninit = octeon3_eth_ndo_uninit,
+ .ndo_open = octeon3_eth_bgx_ndo_open,
+ .ndo_stop = octeon3_eth_bgx_ndo_stop,
+ .ndo_start_xmit = octeon3_eth_ndo_start_xmit,
+ .ndo_get_stats64 = octeon3_eth_ndo_get_stats64,
+ .ndo_set_rx_mode = bgx_port_set_rx_filtering,
+ .ndo_set_mac_address = octeon3_eth_set_mac_address,
+ .ndo_change_mtu = octeon3_eth_ndo_change_mtu,
+ .ndo_do_ioctl = octeon3_ioctl,
+};
+
+static int octeon3_eth_probe(struct platform_device *pdev)
+{
+ struct octeon3_ethernet *priv;
+ struct net_device *netdev;
+ int r;
+
+ struct mac_platform_data *pd = dev_get_platdata(&pdev->dev);
+
+ r = octeon3_eth_global_init(pd->numa_node, pdev);
+ if (r)
+ return r;
+
+ dev_info(&pdev->dev, "Probing %d-%d:%d\n", pd->numa_node, pd->interface,
+ pd->port);
+ netdev = alloc_etherdev(sizeof(struct octeon3_ethernet));
+ if (!netdev) {
+ dev_err(&pdev->dev, "Failed to allocated ethernet device\n");
+ return -ENOMEM;
+ }
+
+ /* Using transmit queues degrades performance significantly */
+ if (!use_tx_queues)
+ netdev->tx_queue_len = 0;
+
+ SET_NETDEV_DEV(netdev, &pdev->dev);
+ dev_set_drvdata(&pdev->dev, netdev);
+
+ if (pd->mac_type == BGX_MAC)
+ bgx_port_set_netdev(pdev->dev.parent, netdev);
+ priv = netdev_priv(netdev);
+ priv->netdev = netdev;
+ priv->mac_type = pd->mac_type;
+ INIT_LIST_HEAD(&priv->list);
+ priv->node = pd->numa_node;
+
+ mutex_lock(&octeon3_eth_node[priv->node].device_list_lock);
+ list_add_tail_rcu(&priv->list,
+ &octeon3_eth_node[priv->node].device_list);
+ mutex_unlock(&octeon3_eth_node[priv->node].device_list_lock);
+
+ priv->index = pd->port;
+ priv->interface = pd->interface;
+ spin_lock_init(&priv->stat_lock);
+
+ if (pd->src_type == XCV)
+ snprintf(netdev->name, IFNAMSIZ, "rgmii%d", pd->port);
+
+ if (priv->mac_type == BGX_MAC)
+ netdev->netdev_ops = &octeon3_eth_netdev_ops;
+
+ if (register_netdev(netdev) < 0) {
+ dev_err(&pdev->dev, "Failed to register ethernet device\n");
+ list_del(&priv->list);
+ free_netdev(netdev);
+ }
+
+ spin_lock_init(&priv->ptp_lock);
+ priv->ptp_info.owner = THIS_MODULE;
+ snprintf(priv->ptp_info.name, 16, "octeon3 ptp");
+ priv->ptp_info.max_adj = 250000000;
+ priv->ptp_info.n_alarm = 0;
+ priv->ptp_info.n_ext_ts = 0;
+ priv->ptp_info.n_per_out = 0;
+ priv->ptp_info.pps = 0;
+ priv->ptp_info.adjfreq = octeon3_adjfreq;
+ priv->ptp_info.adjtime = octeon3_adjtime;
+ priv->ptp_info.gettime64 = octeon3_gettime;
+ priv->ptp_info.settime64 = octeon3_settime;
+ priv->ptp_info.enable = octeon3_enable;
+ priv->ptp_clock = ptp_clock_register(&priv->ptp_info, &pdev->dev);
+
+ netdev_info(netdev, "Registered\n");
+ return 0;
+}
+
+/* octeon3_eth_global_exit - Free all the used resources and restore the
+ * hardware to the default state.
+ * @node: Node to free/reset.
+ *
+ * Returns 0 if successful.
+ * Returns <0 for error codes.
+ */
+static int octeon3_eth_global_exit(int node)
+{
+ struct octeon3_ethernet_node *oen = octeon3_eth_node + node;
+ int i;
+
+ /* Free the tx_complete irq */
+ octeon3_sso_irq_set(node, oen->tx_complete_grp, false);
+ irq_set_affinity_hint(oen->tx_irq, NULL);
+ free_irq(oen->tx_irq, oen);
+ irq_dispose_mapping(oen->tx_irq);
+ oen->tx_irq = 0;
+
+ /* Stop the worker threads */
+ for (i = 0; i < ARRAY_SIZE(oen->workers); i++)
+ kthread_stop(oen->workers[i].task);
+
+ /* Shutdown pki */
+ octeon3_pki_shutdown(node);
+ octeon_fpa3_release_pool(node, oen->pki_packet_pool);
+ kfree(oen->pki_packet_pool_stack);
+
+ /* Shutdown pko */
+ octeon3_pko_exit_global(node);
+ for (;;) {
+ void **w;
+
+ w = octeon_fpa3_alloc(node, oen->pko_aura);
+ if (!w)
+ break;
+ kmem_cache_free(octeon3_eth_sso_pko_cache, w);
+ }
+ octeon_fpa3_release_aura(node, oen->pko_aura);
+ octeon_fpa3_release_pool(node, oen->pko_pool);
+ kfree(oen->pko_pool_stack);
+
+ /* Shutdown sso */
+ octeon3_sso_shutdown(node, oen->sso_aura);
+ octeon3_sso_free_groups(node, &oen->tx_complete_grp, 1);
+ for (;;) {
+ void **w;
+
+ w = octeon_fpa3_alloc(node, oen->sso_aura);
+ if (!w)
+ break;
+ kmem_cache_free(octeon3_eth_sso_pko_cache, w);
+ }
+ octeon_fpa3_release_aura(node, oen->sso_aura);
+ octeon_fpa3_release_pool(node, oen->sso_pool);
+ kfree(oen->sso_pool_stack);
+
+ return 0;
+}
+
+static int octeon3_eth_remove(struct platform_device *pdev)
+{
+ struct mac_platform_data *pd = dev_get_platdata(&pdev->dev);
+ struct net_device *netdev = dev_get_drvdata(&pdev->dev);
+ struct octeon3_ethernet *priv = netdev_priv(netdev);
+ struct octeon3_ethernet_node *oen;
+ int node = priv->node;
+
+ oen = octeon3_eth_node + node;
+
+ ptp_clock_unregister(priv->ptp_clock);
+ unregister_netdev(netdev);
+ if (pd->mac_type == BGX_MAC)
+ bgx_port_set_netdev(pdev->dev.parent, NULL);
+ dev_set_drvdata(&pdev->dev, NULL);
+
+ /* Free all resources when there are no more devices */
+ mutex_lock(&octeon3_eth_init_mutex);
+ mutex_lock(&oen->device_list_lock);
+ list_del_rcu(&priv->list);
+ if (oen->init_done && list_empty(&oen->device_list)) {
+ int i;
+
+ for (i = 0; i < MAX_NAPIS_PER_NODE; i++) {
+ napi_disable(&napi_wrapper[node][i].napi);
+ netif_napi_del(&napi_wrapper[node][i].napi);
+ }
+
+ oen->init_done = false;
+ oen->napi_init_done = false;
+ octeon3_eth_global_exit(node);
+ }
+
+ mutex_unlock(&oen->device_list_lock);
+ mutex_unlock(&octeon3_eth_init_mutex);
+ free_netdev(netdev);
+
+ return 0;
+}
+
+static void octeon3_eth_shutdown(struct platform_device *pdev)
+{
+ octeon3_eth_remove(pdev);
+}
+
+static struct platform_driver octeon3_eth_driver = {
+ .probe = octeon3_eth_probe,
+ .remove = octeon3_eth_remove,
+ .shutdown = octeon3_eth_shutdown,
+ .driver = {
+ .owner = THIS_MODULE,
+ .name = "ethernet-mac-pki",
+ },
+};
+
+static int __init octeon3_eth_init(void)
+{
+ if (rx_contexts <= 0)
+ rx_contexts = 1;
+ if (rx_contexts > MAX_RX_CONTEXTS)
+ rx_contexts = MAX_RX_CONTEXTS;
+
+ return platform_driver_register(&octeon3_eth_driver);
+}
+module_init(octeon3_eth_init);
+
+static void __exit octeon3_eth_exit(void)
+{
+ platform_driver_unregister(&octeon3_eth_driver);
+
+ /* Destroy the memory cache used by sso and pko */
+ kmem_cache_destroy(octeon3_eth_sso_pko_cache);
+}
+module_exit(octeon3_eth_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Cavium, Inc. <support@caviumnetworks.com>");
+MODULE_DESCRIPTION("Cavium, Inc. PKI/PKO Ethernet driver.");
--
2.1.4
^ permalink raw reply related
* [PATCH net-next v11 07/10] netdev: cavium: octeon: Add Octeon III SSO Support
From: Steven J. Hill @ 2018-06-04 22:00 UTC (permalink / raw)
To: netdev; +Cc: Carlos Munoz, Steven J. Hill
In-Reply-To: <1528149617-8964-1-git-send-email-steven.hill@cavium.com>
From: Carlos Munoz <cmunoz@cavium.com>
Add support for Octeon III SSO logic block for BGX Ethernet.
Signed-off-by: Carlos Munoz <cmunoz@cavium.com>
Signed-off-by: Steven J. Hill <Steven.Hill@cavium.com>
---
drivers/net/ethernet/cavium/octeon/octeon3-sso.c | 244 +++++++++++++++++++++++
1 file changed, 244 insertions(+)
create mode 100644 drivers/net/ethernet/cavium/octeon/octeon3-sso.c
diff --git a/drivers/net/ethernet/cavium/octeon/octeon3-sso.c b/drivers/net/ethernet/cavium/octeon/octeon3-sso.c
new file mode 100644
index 0000000..51d67a8
--- /dev/null
+++ b/drivers/net/ethernet/cavium/octeon/octeon3-sso.c
@@ -0,0 +1,244 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Octeon III Schedule/Synchronize/Order Unit (SSO)
+ *
+ * Copyright (C) 2018 Cavium, Inc.
+ */
+
+#include "octeon3.h"
+
+/* Registers are accessed via xkphys. */
+#define SSO_BASE 0x1670000000000ull
+#define SSO_ADDR(node) (SET_XKPHYS + NODE_OFFSET(node) + SSO_BASE)
+
+#define SSO_AW_STATUS(n) (SSO_ADDR(n) + 0x000010e0)
+#define SSO_AW_CFG(n) (SSO_ADDR(n) + 0x000010f0)
+#define SSO_ERR0(n) (SSO_ADDR(n) + 0x00001240)
+#define SSO_TAQ_ADD(n) (SSO_ADDR(n) + 0x000020e0)
+#define SSO_XAQ_AURA(n) (SSO_ADDR(n) + 0x00002100)
+
+#define AQ_OFFSET(g) ((g) << 3)
+#define AQ_ADDR(n, g) (SSO_ADDR(n) + AQ_OFFSET(g))
+#define SSO_XAQ_HEAD_PTR(n, g) (AQ_ADDR(n, g) + 0x00080000)
+#define SSO_XAQ_TAIL_PTR(n, g) (AQ_ADDR(n, g) + 0x00090000)
+#define SSO_XAQ_HEAD_NEXT(n, g) (AQ_ADDR(n, g) + 0x000a0000)
+#define SSO_XAQ_TAIL_NEXT(n, g) (AQ_ADDR(n, g) + 0x000b0000)
+
+#define GRP_OFFSET(grp) ((grp) << 16)
+#define GRP_ADDR(n, g) (SSO_ADDR(n) + GRP_OFFSET(g))
+#define SSO_GRP_TAQ_THR(n, g) (GRP_ADDR(n, g) + 0x20000100)
+#define SSO_GRP_PRI(n, g) (GRP_ADDR(n, g) + 0x20000200)
+#define SSO_GRP_INT(n, g) (GRP_ADDR(n, g) + 0x20000400)
+#define SSO_GRP_INT_THR(n, g) (GRP_ADDR(n, g) + 0x20000500)
+#define SSO_GRP_AQ_CNT(n, g) (GRP_ADDR(n, g) + 0x20000700)
+
+static int octeon3_sso_get_num_groups(void)
+{
+ if (OCTEON_IS_MODEL(OCTEON_CN78XX))
+ return 256;
+ if (OCTEON_IS_MODEL(OCTEON_CNF75XX) || OCTEON_IS_MODEL(OCTEON_CN73XX))
+ return 64;
+ return 0;
+}
+
+void octeon3_sso_irq_set(int node, int group, bool enable)
+{
+ if (enable)
+ oct_csr_write(1, SSO_GRP_INT_THR(node, group));
+ else
+ oct_csr_write(0, SSO_GRP_INT_THR(node, group));
+
+ oct_csr_write(BIT(1), SSO_GRP_INT(node, group));
+}
+EXPORT_SYMBOL(octeon3_sso_irq_set);
+
+/* octeon3_sso_alloc_groups - Allocate a range of SSO groups.
+ * @node: Node where SSO resides.
+ * @groups: Pointer to allocated groups.
+ * @cnt: Number of groups to allocate.
+ * @start: Group number to start sequential allocation from. -1 for don't care.
+ *
+ * Returns 0 if successful, error code otherwise..
+ */
+int octeon3_sso_alloc_groups(int node, int *groups, int cnt, int start)
+{
+ struct global_resource_tag tag;
+ int group, ret;
+ char buf[16];
+
+ strncpy((char *)&tag.lo, "cvm_sso_", 8);
+ snprintf(buf, 16, "0%d......", node);
+ memcpy(&tag.hi, buf, 8);
+
+ res_mgr_create_resource(tag, octeon3_sso_get_num_groups());
+
+ if (!groups)
+ ret = res_mgr_alloc_range(tag, start, cnt, false, &group);
+ if (!ret)
+ ret = group;
+ else
+ ret = res_mgr_alloc_range(tag, start, cnt, false, groups);
+
+ return ret;
+}
+EXPORT_SYMBOL(octeon3_sso_alloc_groups);
+
+/* octeon3_sso_free_groups - Free SSO groups.
+ * @node: Node where SSO resides.
+ * @groups: Array of groups to free.
+ * @cnt: Number of groups to free.
+ */
+void octeon3_sso_free_groups(int node, int *groups, int cnt)
+{
+ struct global_resource_tag tag;
+ char buf[16];
+
+ /* Allocate the requested groups. */
+ strncpy((char *)&tag.lo, "cvm_sso_", 8);
+ snprintf(buf, 16, "0%d......", node);
+ memcpy(&tag.hi, buf, 8);
+
+ res_mgr_free_range(tag, groups, cnt);
+}
+EXPORT_SYMBOL(octeon3_sso_free_groups);
+
+/* octeon3_sso_pass1_limit - When the Transitory Admission Queue (TAQ) is
+ * almost full, it is possible for the SSo to hang. We work around this
+ * by ensuring that the sum of SSO_GRP(0..255)_TAQ_THR[MAX_THR] of all
+ * used groups is <= 1264. This may reduce single group performance when
+ * many groups are in use.
+ * @node: Node to update.
+ * @grp: SSO group to update.
+ */
+void octeon3_sso_pass1_limit(int node, int group)
+{
+ u64 max_thr, rsvd_thr, taq_add, taq_thr;
+
+ /* Ideally we would like to divide the maximum number of TAQ buffers
+ * (1264) among the SSO groups in use. However, since we do not know
+ * how many SSO groups are used by code outside this driver, we take
+ * the worst case approach.
+ */
+ max_thr = 1264 / octeon3_sso_get_num_groups();
+ if (max_thr < 4)
+ max_thr = 4;
+ rsvd_thr = max_thr - 1;
+
+ /* Changes to SSO_GRP_TAQ_THR[rsvd_thr] must also update
+ * SSO_TAQ_ADD[RSVD_FREE].
+ */
+ taq_thr = oct_csr_read(SSO_GRP_TAQ_THR(node, group));
+ taq_add = (rsvd_thr - (taq_thr & GENMASK_ULL(10, 0))) << 16;
+
+ taq_thr &= ~(GENMASK_ULL(42, 32) | GENMASK_ULL(10, 0));
+ taq_thr |= max_thr << 32;
+ taq_thr |= rsvd_thr;
+
+ oct_csr_write(taq_thr, SSO_GRP_TAQ_THR(node, group));
+ oct_csr_write(taq_add, SSO_TAQ_ADD(node));
+}
+EXPORT_SYMBOL(octeon3_sso_pass1_limit);
+
+/* octeon3_sso_shutdown - Shutdown the SSO.
+ * @node: Node where SSO to disable is.
+ * @aura: Aura used for the SSO buffers.
+ */
+void octeon3_sso_shutdown(int node, int aura)
+{
+ int i, max_grps, timeout;
+ u64 data, head, tail;
+ void *ptr;
+
+ /* Disable SSO. */
+ data = oct_csr_read(SSO_AW_CFG(node));
+ data |= BIT(6) | BIT(4);
+ data &= ~BIT(0);
+ oct_csr_write(data, SSO_AW_CFG(node));
+
+ /* Extract the FPA buffers. */
+ max_grps = octeon3_sso_get_num_groups();
+ for (i = 0; i < max_grps; i++) {
+ head = oct_csr_read(SSO_XAQ_HEAD_PTR(node, i));
+ tail = oct_csr_read(SSO_XAQ_TAIL_PTR(node, i));
+ data = oct_csr_read(SSO_GRP_AQ_CNT(node, i));
+
+ /* Verify pointers. */
+ head &= GENMASK_ULL(41, 7);
+ tail &= GENMASK_ULL(41, 7);
+ if (head != tail) {
+ pr_err("%s: Bad pointer\n", __func__);
+ continue;
+ }
+
+ /* This SSO group should have no pending entries. */
+ if (data & GENMASK_ULL(32, 0))
+ pr_err("%s: Group not empty\n", __func__);
+
+ ptr = phys_to_virt(head);
+ octeon_fpa3_free(node, aura, ptr);
+
+ /* Clear pointers. */
+ oct_csr_write(0, SSO_XAQ_HEAD_PTR(node, i));
+ oct_csr_write(0, SSO_XAQ_HEAD_NEXT(node, i));
+ oct_csr_write(0, SSO_XAQ_TAIL_PTR(node, i));
+ oct_csr_write(0, SSO_XAQ_TAIL_NEXT(node, i));
+ }
+
+ /* Make sure all buffers are drained. */
+ timeout = 10000;
+ do {
+ data = oct_csr_read(SSO_AW_STATUS(node));
+ if ((data & GENMASK_ULL(5, 0)) == 0)
+ break;
+ timeout--;
+ udelay(1);
+ } while (timeout);
+ if (!timeout)
+ pr_err("%s: Timed out draining buffers\n", __func__);
+}
+EXPORT_SYMBOL(octeon3_sso_shutdown);
+
+/* octeon3_sso_init - Initialize the SSO.
+ * @node: Node where SSO resides.
+ * @aura: Aura used for the SSO buffers.
+ */
+int octeon3_sso_init(int node, int aura)
+{
+ int i, max_grps;
+ u64 data, phys;
+ int err = 0;
+ void *mem;
+
+ data = BIT(3) | BIT(2) | BIT(1);
+ oct_csr_write(data, SSO_AW_CFG(node));
+
+ data = (node << 10) | aura;
+ oct_csr_write(data, SSO_XAQ_AURA(node));
+
+ max_grps = octeon3_sso_get_num_groups();
+ for (i = 0; i < max_grps; i++) {
+ mem = octeon_fpa3_alloc(node, aura);
+ if (!mem) {
+ err = -ENOMEM;
+ goto out;
+ }
+
+ phys = virt_to_phys(mem);
+ oct_csr_write(phys, SSO_XAQ_HEAD_PTR(node, i));
+ oct_csr_write(phys, SSO_XAQ_HEAD_NEXT(node, i));
+ oct_csr_write(phys, SSO_XAQ_TAIL_PTR(node, i));
+ oct_csr_write(phys, SSO_XAQ_TAIL_NEXT(node, i));
+
+ /* SSO-18678 */
+ data = 0x3f << 16;
+ oct_csr_write(data, SSO_GRP_PRI(node, i));
+ }
+
+ data = BIT(0);
+ oct_csr_write(data, SSO_ERR0(node));
+
+ data = BIT(3) | BIT(2) | BIT(1) | BIT(0);
+ oct_csr_write(data, SSO_AW_CFG(node));
+out:
+ return err;
+}
+EXPORT_SYMBOL(octeon3_sso_init);
--
2.1.4
^ 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