Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 9 May 2018 11:00:52 +1000
From:      Lawrence Stewart <lstewart@freebsd.org>
To:        Tom Jones <tj@enoti.me>, Harsh Jain <harsh@chelsio.com>
Cc:        freebsd-net@freebsd.org, Navdeep Parhar <navdeep@chelsio.com>, John Baldwin <jhb@freebsd.org>, sonyarpitad@chelsio.com
Subject:   Re: Bug: Newreno; Seems Memory leak in newreno_cb_init
Message-ID:  <6c84c94a-9d16-db30-f8d5-c6b364cd9469@freebsd.org>
In-Reply-To: <20180508200439.GA32339@tom-desk.erg.abdn.ac.uk>
References:  <58ab6c13-f9c4-b4d7-7f2c-eade3749457f@chelsio.com> <20180508200439.GA32339@tom-desk.erg.abdn.ac.uk>

next in thread | previous in thread | raw e-mail | index | archive | help
On 09/05/2018 06:04, Tom Jones wrote:
> On Tue, May 08, 2018 at 05:14:49PM +0530, Harsh Jain wrote:
>> Hi All,
>>
>> We have observed memory leak with TCP network traffic in "newreno".
>>
>> Output of vmstat -m
>>
>>    in_mfilter     3     3K       -        3  1024
>>      in_multi     4     1K       -        4  256
>>   ip_moptions     6     1K       -        6  64,256
>> encap_export_host     2     2K       -        2  1024
>>  newreno data 394849273 6169520K       - 394849273  16
>>     sctp_a_it     0     0K       -        5  16
>>      sctp_vrf     1     1K       -        1  64
>>      sctp_ifa     7     1K       -        7  128
>>      sctp_ifn     4     1K       -        4  128
>>
>> There is 1 malloc in "newreno_cb_init" whose pointer is not saved in any global structure to free the same.
>>
>> Is this a BUG?
> 
> 
> Hi Harsh,
> 
> Adding Lawrence in cc
> 
> It looks like it, running nc in a loop I can watch MemUse grow. 
> 
> I think this should address the leak
> 
> https://reviews.freebsd.org/D15358

I'm not clear why yet, but the patch I ultimately committed as r331214
is deeply flawed on account of missing memory allocation and other
changes that never ended up in the working copy I committed from. The
cb_destroy() change for example exists in the D11616 Phabricator review
though. I think I may have refined the final patch and committed from a
working copy that started with an older stale version of the patch. Ugh.

Mea culpa, thanks for the bug report, and apologies for the oversight.
Will work with Tom to get this fixed.

Cheers,
Lawrence



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?6c84c94a-9d16-db30-f8d5-c6b364cd9469>