On Fri, Dec 04, 2020 at 07:42:56AM +0200, Julian Anastasov wrote:
>
> Hello,
>
>On Fri, 4 Dec 2020, dust.li wrote:
>
>>
>> On 12/3/20 4:48 PM, Julian Anastasov wrote:
>> >
>> > - work will use spin_lock_bh(&s->lock) to protect the
>&
On 12/3/20 4:48 PM, Julian Anastasov wrote:
Hello,
On Thu, 3 Dec 2020, dust.li wrote:
Hi Yunhong & Julian, any updates ?
We've encountered the same problem. With lots of ipvs
services plus many CPUs, it's easy to reproduce this issue.
I have a simple script to rep
Hi Yunhong & Julian, any updates ?
We've encountered the same problem. With lots of ipvs
services plus many CPUs, it's easy to reproduce this issue.
I have a simple script to reproduce:
First add many ipvs services:
for((i=0;i<5;i++)); do
ipvsadm -A -t 10.10.10.10:$((2000+$i))
do
ubmit() won't do string formatting
but just pass them to the userspace perf, which will finnally did the
formatting, but it's already too late to get sysctl_mem[x].
So, any pointer dereference in tracepoint entry should failed in
`perf trace`.
Thanks.
Dust.Li
On Tue, Sep 08, 2020 at 10:46:13AM +0200, Paolo Abeni wrote:
>Hi,
>
>On Tue, 2020-09-08 at 11:15 +0800, dust.li wrote:
>> Actually, with more udp sockets, I can always make it large
>> enough to exceed udp_mem[0], and drop packets before udp_mem[1]
>> and udp_mem[2
On Mon, Sep 07, 2020 at 07:18:48PM +0200, Paolo Abeni wrote:
>Hi,
>
>On Mon, 2020-09-07 at 22:44 +0800, Dust Li wrote:
>> We encoutered udp packets drop under a pretty low pressure
>> with net.ipv4.udp_mem[0] set to a small value (4096).
>>
>> After some tracing and debugging, we found that for ud
On Mon, Sep 07, 2020 at 10:50:30AM -0700, Jakub Kicinski wrote:
>On Mon, 7 Sep 2020 22:47:57 +0800 Dust Li wrote:
>> sysctl_mem is an point, and tracepoint entry do not support
>> been visited like an array. Use 3 long type to get sysctl_mem
>> instead.
>>
>> tracpoint output with and without thi
