On Wed 23 May 2018 at 23:14, Cong Wang wrote:
> On Wed, May 23, 2018 at 1:52 AM, Vlad Buslov wrote:
>> Initial net_device implementation used ingress_lock spinlock to synchronize
>> ingress path of device. This lock was used in both process and bh context.
>> In some code
On Thu 24 May 2018 at 23:34, Cong Wang wrote:
> On Mon, May 14, 2018 at 7:27 AM, Vlad Buslov wrote:
>> Currently, all netlink protocol handlers for updating rules, actions and
>> qdiscs are protected with single global rtnl lock which removes any
>> possibility for parall
Retry check-insert sequence in action init functions if action with same
index was inserted concurrently.
Signed-off-by: Vlad Buslov
---
net/sched/act_bpf.c| 8 +++-
net/sched/act_connmark.c | 8 +++-
net/sched/act_csum.c | 8 +++-
net/sched/act_gact.c | 8
Implement functions to atomically update and free action cookie
using rcu mechanism.
Signed-off-by: Vlad Buslov
---
include/net/act_api.h | 2 +-
include/net/pkt_cls.h | 1 +
net/sched/act_api.c | 44 ++--
3 files changed, 32 insertions(+), 15
Implement delete function that is required to delete actions without
holding rtnl lock. Use action API function that atomically deletes action
only if it is still in action idr. This implementation prevents concurrent
threads from deleting same action twice.
Signed-off-by: Vlad Buslov
---
net
'delete' function and 'unlocked' flag.
- Change action API to work with actions in lockless manner based on
primitives implemented in previous patches.
- Extend action API with new functions necessary to implement unlocked
actions.
Vlad Buslov (14):
net: sched: use rcu for
Implement new action API function to atomically delete action with
specified index and to atomically insert unique action. These functions are
required to implement init and delete functions for specific actions that
do not rely on rtnl lock.
Signed-off-by: Vlad Buslov
---
include/net/act_api.h
Extend rate estimator new and replace APIs with additional spinlock
parameter used by lockless actions to protect rate_est pointer from
concurrent modification.
Signed-off-by: Vlad Buslov
---
include/net/gen_stats.h| 2 ++
net/core/gen_estimator.c | 58
Substitute calls to action insert function with calls to action insert
unique function that warns if insertion overwrites index in idr.
Signed-off-by: Vlad Buslov
---
net/sched/act_bpf.c| 2 +-
net/sched/act_connmark.c | 2 +-
net/sched/act_csum.c | 2 +-
net/sched/act_gact.c
Add additional 'unlocked' argument to act API init functions.
Argument is true when rtnl lock is not taken and false otherwise.
It is required to implement actions that need to release rtnl lock before
loading kernel module and reacquire if afterwards.
Signed-off-by: Vlad Buslov
---
i
rtnl lock. Remove rtnl lock assertions that are no longer
needed.
Signed-off-by: Vlad Buslov
---
net/sched/act_api.c | 98 -
net/sched/cls_api.c | 1 -
2 files changed, 82 insertions(+), 17 deletions(-)
diff --git a/net/sched/act_api.c b/net
orary reference and not expose it to userspace.
Signed-off-by: Vlad Buslov
---
net/sched/act_api.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/net/sched/act_api.c b/net/sched/act_api.c
index 3f02cd1..2772276e 100644
--- a/net/sched/act_api.c
+++ b/net/sc
user must always hold reference to it.
Implement helper put list function to atomically release list of actions
after action API init code is done using them.
Signed-off-by: Vlad Buslov
---
net/sched/act_api.c | 38 +-
1 file changed, 17 insertions(+), 21
Change type of action reference counter to refcount_t.
Change type of action bind counter to atomic_t.
This type is used to allow decrementing bind counter without testing
for 0 result.
Signed-off-by: Vlad Buslov
---
include/net/act_api.h | 5 +++--
net/sched/act_api.c| 32
always take reference to action before returning
successfully.
Signed-off-by: Vlad Buslov
---
net/sched/act_bpf.c| 8
net/sched/act_connmark.c | 5 +++--
net/sched/act_csum.c | 8
net/sched/act_gact.c | 5 +++--
net/sched/act_ife.c| 12
its reference and bind counters.
Implement both action search and check using new safe function.
Signed-off-by: Vlad Buslov
---
net/sched/act_api.c | 38 --
1 file changed, 16 insertions(+), 22 deletions(-)
diff --git a/net/sched/act_api.c b/net/sched/act_api.c
Extend action ops with 'delete' function. Each action type to implement its
own delete function that doesn't depend on rtnl lock.
Signed-off-by: Vlad Buslov
---
include/net/act_api.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/include/net/act_api.h b/include/net/act_api.
gt;Remove unsafe action idr lookup function. Instead of it, implement safe tcf
>>idr check function that atomically looks up action in idr and increments
>>its reference and bind counters.
>>
>>Implement both action search and check using new safe function.
>>
>&g
On Mon 14 May 2018 at 16:47, Jiri Pirko wrote:
> Mon, May 14, 2018 at 04:27:07PM CEST, vla...@mellanox.com wrote:
>
> [...]
>
>
>>+static int tcf_action_del_1(struct net *net, char *kind, u32 index,
>>+ struct netlink_ext_ack *extack)
>>+{
>>+ const struct tc_action_op
On Mon 14 May 2018 at 18:03, Jamal Hadi Salim wrote:
> On 14/05/18 10:27 AM, Vlad Buslov wrote:
>> Currently, all netlink protocol handlers for updating rules, actions and
>> qdiscs are protected with single global rtnl lock which removes any
>> possibility for parallelism
Underlying implementation of action map has changed and doesn't require
disabling bh anymore. Replace all action idr spinlock usage with regular
calls that do not disable bh.
Signed-off-by: Vlad Buslov
---
net/sched/act_api.c | 20 ++--
1 file changed, 10 insertions(+
Test 6fb4 creates one mirred and one pipe action, but only flushes mirred
on teardown. Leaking pipe action causes failures in other tests.
Add additional teardown command to also flush gact actions.
Signed-off-by: Vlad Buslov
---
tools/testing/selftests/tc-testing/tc-tests/actions/mirred.json
Fix expected ip address to actually match configured ip address.
Fix test to expect single matched filter.
Signed-off-by: Vlad Buslov
---
tools/testing/selftests/tc-testing/tc-tests/filters/tests.json | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/tools/testing
On Tue 16 Feb 2021 at 01:22, Cong Wang wrote:
> On Wed, Feb 10, 2021 at 9:53 PM syzbot
> wrote:
>>
>> syzbot has found a reproducer for the following issue on:
>>
>> HEAD commit:291009f6 Merge tag 'pm-5.11-rc8' of git://git.kernel.org/p..
>> git tree: upstream
>> console output: https:/
ction insertions together")
Reported-by: syzbot+151e3e714d34ae4ce...@syzkaller.appspotmail.com
Signed-off-by: Vlad Buslov
---
include/net/act_api.h | 1 +
net/sched/act_api.c | 2 +-
net/sched/cls_api.c | 1 +
3 files changed, 3 insertions(+), 1 deletion(-)
diff --git a/include/net/act_api.h b/i
On Wed 23 May 2018 at 01:10, Cong Wang wrote:
> On Mon, May 21, 2018 at 1:03 PM, Vlad Buslov wrote:
>> Initial net_device implementation used ingress_lock spinlock to synchronize
>> ingress path of device. This lock was used in both process and bh context.
>> In some code
eplace all action idr spinlock usage with regular calls that do not
disable bh.
Acked-by: Jiri Pirko
Acked-by: Jamal Hadi Salim
Signed-off-by: Vlad Buslov
---
Changes from V2 to V3:
- Expanded commit message.
Changes from V1 to V2:
- Expanded commit message.
net/sched/act_api.c | 20 ++
On Tue 15 May 2018 at 09:03, Jiri Pirko wrote:
> Mon, May 14, 2018 at 09:07:06PM CEST, vla...@mellanox.com wrote:
>>
>>On Mon 14 May 2018 at 16:47, Jiri Pirko wrote:
>>> Mon, May 14, 2018 at 04:27:07PM CEST, vla...@mellanox.com wrote:
>>>
>>> [...]
>>>
>>>
+static int tcf_action_del_1(struct
t;>after action API init code is done using them.
>>
>>Signed-off-by: Vlad Buslov
>>---
>> net/sched/act_api.c | 38 +-
>> 1 file changed, 17 insertions(+), 21 deletions(-)
>>
>
> [...]
>
>
>>@@ -1196,8 +1190,
or concurrent
>>>>modifications so user must always hold reference to it.
>>>>
>>>>Implement helper put list function to atomically release list of actions
>>>>after action API init code is done using them.
>>>>
>>>>Signed-off-b
atomically looks up action in idr and increments
>>>>its reference and bind counters.
>>>>
>>>>Implement both action search and check using new safe function.
>>>>
>>>>Signed-off-by: Vlad Buslov
>>>>---
>>>> net/sched/act_
>>>
>>>>>>Implement helper put list function to atomically release list of actions
>>>>>>after action API init code is done using them.
>>>>>>
>>>>>>Signed-off-by: Vlad Buslov
>>>>>>---
>>
On Tue 15 May 2018 at 18:25, Jamal Hadi Salim wrote:
> On 14/05/18 04:46 PM, Vlad Buslov wrote:
>>
>> On Mon 14 May 2018 at 18:03, Jamal Hadi Salim wrote:
>>> On 14/05/18 10:27 AM, Vlad Buslov wrote:
>
>
>> Hello Jamal,
>>
>> I'm trying
ence to action before returning
>>successfully.
>
> Where's the balance? Who does the release instead? I'm probably missing
> something.
I've resplit these patches for V2 to always do take/release in same
patch.
>
>>
>>Signed-off-by: Vlad Buslov
to implement init and delete functions for specific actions that
>>do not rely on rtnl lock.
>>
>>Signed-off-by: Vlad Buslov
>>---
>> include/net/act_api.h | 2 ++
>> net/sched/act_api.c | 45 +
>> 2 files chan
on API function to atomically delete action with
>>>>specified index and to atomically insert unique action. These functions are
>>>>required to implement init and delete functions for specific actions that
>>>>do not rely on rtnl lock.
>>>>
>&
On Wed 16 May 2018 at 09:50, Jiri Pirko wrote:
> Mon, May 14, 2018 at 04:27:14PM CEST, vla...@mellanox.com wrote:
>>Substitute calls to action insert function with calls to action insert
>>unique function that warns if insertion overwrites index in idr.
>>
>>
s still in action idr. This implementation prevents concurrent
>>threads from deleting same action twice.
>>
>>Signed-off-by: Vlad Buslov
>>---
>> net/sched/act_bpf.c| 8
>> net/sched/act_connmark.c | 8
>> net/sched/ac
tion.
>>
>>Signed-off-by: Vlad Buslov
>
> [...]
>
>
>> /**
>> * gen_new_estimator - create a new rate estimator
>> * @bstats: basic statistics
>> * @cpu_bstats: bstats per cpu
>> * @rate_est: rate estimator statistics
>>+ * @rate_e
On Wed 16 May 2018 at 09:59, Jiri Pirko wrote:
> Mon, May 14, 2018 at 04:27:13PM CEST, vla...@mellanox.com wrote:
>>Retry check-insert sequence in action init functions if action with same
>>index was inserted concurrently.
>>
>>Signed-off-by: Vlad Buslov
>>---
sequence in action init functions if action with same
>>>>index was inserted concurrently.
>>>>
>>>>Signed-off-by: Vlad Buslov
>>>>---
>>>> net/sched/act_bpf.c| 8 +++-
>>>> net/sched/act_connmark.c | 8 +
t;On Wed 16 May 2018 at 09:59, Jiri Pirko wrote:
>>>>> Mon, May 14, 2018 at 04:27:13PM CEST, vla...@mellanox.com wrote:
>>>>>>Retry check-insert sequence in action init functions if action with same
>>>>>>index was inserted concurrently.
>>>>
.@mellanox.com wrote:
>>>>>>>>Retry check-insert sequence in action init functions if action with same
>>>>>>>>index was inserted concurrently.
>>>>>>>>
>>>>>>>>Signed-off-by: Vlad Buslov
>>&g
On Thu 28 Feb 2019 at 02:18, Dennis Zhou wrote:
> Hi everyone,
>
> It was reported a while [1] that an increase in allocation alignment
> requirement [2] caused the percpu memory allocator to do significantly
> more work.
>
> After spending quite a bit of time diving into it, it seems the crux w
flow_match_level match_level;
> | ^~~
> drivers/net/ethernet/mellanox/mlx5/core/en/tc_tun.h:27:26: warning: 'struct
> mlx5e_encap_entry' declared inside parameter list will not be visible outside
> of this definition or declaration
&g
--git a/net/sched/act_api.c b/net/sched/act_api.c
> index b919826939e0..43cceb924976 100644
> --- a/net/sched/act_api.c
> +++ b/net/sched/act_api.c
> @@ -1042,6 +1042,9 @@ struct tc_action *tcf_action_init_1(struct net *net,
> struct tcf_proto *tp,
> if (err != ACT_P_CREATED)
> module_put(a_o->owner);
>
> + if (!bind && ovr && err == ACT_P_CREATED)
> + refcount_set(&a->tcfa_refcnt, 2);
> +
> return a;
>
> err_out:
Reviewed-and-tested-by: Vlad Buslov
xist is
handled correctly.
Signed-off-by: Vlad Buslov
---
.../tc-testing/tc-tests/actions/simple.json | 24 +++
1 file changed, 24 insertions(+)
diff --git a/tools/testing/selftests/tc-testing/tc-tests/actions/simple.json
b/tools/testing/selftests/tc-testing/tc-tests/actions/si
On Wed 31 Mar 2021 at 07:40, Cong Wang wrote:
> On Mon, Mar 29, 2021 at 3:55 PM Kumar Kartikeya Dwivedi
> wrote:
>> diff --git a/net/sched/act_api.c b/net/sched/act_api.c
>> index b919826939e0..43cceb924976 100644
>> --- a/net/sched/act_api.c
>> +++ b/net/sched/act_api.c
>> @@ -1042,6 +1042,9 @
On Wed 31 Mar 2021 at 11:37, Kumar Kartikeya Dwivedi wrote:
> Currently, action creation using ACT API in replace mode is buggy. When
> invoking for non-existent action index 42,
>
> tc action replace action bpf obj foo.o sec index 42
>
> kernel creates the action, fills up the netlink r
Hi Kumar,
Thanks for the patch!
On Sun 28 Mar 2021 at 09:45, Kumar Kartikeya Dwivedi wrote:
> When creating an action in replace mode, in tcf_action_add, the refcount
> of existing actions is rightly raised during tcf_idr_check_alloc call,
> but for new actions a dummy placeholder entry is creat
On Thu 25 Mar 2021 at 14:00, Kumar Kartikeya Dwivedi wrote:
> This adds functions that wrap the netlink API used for adding,
> manipulating, and removing filters and actions. These functions operate
> directly on the loaded prog's fd, and return a handle to the filter and
> action using an out p
On Mon 29 Mar 2021 at 15:32, Toke Høiland-Jørgensen wrote:
> Vlad Buslov writes:
>
>> On Thu 25 Mar 2021 at 14:00, Kumar Kartikeya Dwivedi
>> wrote:
>>> This adds functions that wrap the netlink API used for adding,
>>> manipulating, and removing filters
struct mlx5e_encap_entry *e)
> { return -EOPNOTSUPP; }
> #endif
> int mlx5e_tc_tun_route_lookup(struct mlx5e_priv *priv,
Thanks Arnd!
Reviewed-by: Vlad Buslov
81db908c581e
>> compiler: clang version 9.0.0 (/home/glider/llvm/clang
>> 80fee25776c2fb61e74c1ecb1a523375c2500b69)
>> syz repro: https://syzkaller.appspot.com/x/repro.syz?x=116c4b1160
>> C reproducer: https://syzkaller.appspot.com/x/repro.c?x=15ff270d600000
>>
>> The bug was bis
On Tue 17 Sep 2019 at 20:03, Cong Wang wrote:
> On Tue, Sep 17, 2019 at 1:27 AM Vlad Buslov wrote:
>> Hi Cong,
>>
>> Don't see why we would need qdisc tree lock while releasing the
>> reference to (or destroying) previous Qdisc. I've skimmed through other
&
Hi Po,
On Tue 28 Apr 2020 at 06:34, Po Liu wrote:
> Introduce a ingress frame gate control flow action.
> Tc gate action does the work like this:
> Assume there is a gate allow specified ingress frames can be passed at
> specific time slot, and be dropped at specific time slot. Tc filter
> choose
On Tue 28 Apr 2020 at 06:34, Po Liu wrote:
> Add the gate action to the flow action entry. Add the gate parameters to
> the tc_setup_flow_action() queueing to the entries of flow_action_entry
> array provide to the driver.
>
> Signed-off-by: Po Liu
> ---
> include/net/flow_offload.h | 10 ++
This is fixed by Dan Carpenter's patch "net: sched: potential NULL
dereference in tcf_block_find()" that was submitted yesterday.
On Mon 18 Feb 2019 at 20:02, Cong Wang wrote:
> (Cc'ing Vlad, please fix it)
>
> On Wed, Feb 13, 2019 at 9:56 AM syzbot
> wrote:
>>
>> Hello,
>>
>> syzbot found the f
On Sat 19 May 2018 at 02:59, Cong Wang wrote:
> On Fri, May 18, 2018 at 8:45 AM, Vlad Buslov wrote:
>> Underlying implementation of action map has changed and doesn't require
>> disabling bh anymore. Replace all action idr spinlock usage with regular
>> calls that do
On Sat 19 May 2018 at 21:04, Marcelo Ricardo Leitner
wrote:
> On Mon, May 14, 2018 at 05:27:03PM +0300, Vlad Buslov wrote:
>> Change type of action reference counter to refcount_t.
>>
>> Change type of action bind counter to atomic_t.
>> This type is used to allo
On Sun 20 May 2018 at 06:22, Jiri Pirko wrote:
> Sat, May 19, 2018 at 11:43:27PM CEST, marcelo.leit...@gmail.com wrote:
>>On Mon, May 14, 2018 at 05:27:07PM +0300, Vlad Buslov wrote:
>>...
>>> @@ -1052,6 +1088,36 @@ static int tca_action_flush(struct net *net,
On Sat 19 May 2018 at 21:52, Marcelo Ricardo Leitner
wrote:
> On Mon, May 14, 2018 at 05:27:10PM +0300, Vlad Buslov wrote:
>> Return from action init function with reference to action taken,
>> even when overwriting existing action.
>
> Isn't this patch necessary be
ls that do not
disable bh.
Signed-off-by: Vlad Buslov
---
net/sched/act_api.c | 20 ++--
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/net/sched/act_api.c b/net/sched/act_api.c
index 72251241665a..3f4cf930f809 100644
--- a/net/sched/act_api.c
+++ b/net/sched
On Thu 03 Sep 2020 at 18:33, syzbot
wrote:
> syzbot has bisected this issue to:
>
> commit 4e8ddd7f1758ca4ddd0c1f7cf3e66fce736241d2
> Author: Vlad Buslov
> Date: Thu Jul 5 14:24:30 2018 +
>
> net: sched: don't release reference on action overwrite
uce dropped frames which is necessary for user. Status
> update shows how many filtered packets increasing and how many dropped
> in those packets.
>
> v2: Changes
> - Update commit comments suggest by Jiri Pirko.
>
> Signed-off-by: Po Liu
> ---
Reviewed-by: Vlad Buslov
65 matches
Mail list logo