[Swan-dev] leaks and cores

[Andrew Cagney]

Another update:

We're down to:

https://testing.libreswan.org/v4.3-86-g701b488bb9-main/ikev2-59-multiple-acquires-alias/OUTPUT/east.pluto.log.gz
Mar  2 15:56:47.592566: leak: heap logger, item size: 64
Mar  2 15:56:47.592656: leak: heap logger prefix, item size: 26
Mar  2 15:56:47.592737: leak: DH shared secret, item size: 104
Mar  2 15:56:47.592807: leak: DH crypto, item size: 384
Mar  2 15:56:47.592876: leak: dh, item size: 40
Mar  2 15:56:47.592937: leak: calc_dh_local_secret, item size: 32
Mar  2 15:56:47.593002: leak detective found 6 leaks, total size 650

but it only happens occasionally so presumably a race between revival,
crypto helpers, and the main thread during shutdown

On Tue, 19 Jan 2021 at 14:26, Andrew Cagney <andrew.cagney at gmail.com> wrote:
>
> Another update: https://testing.libreswan.org/v4.1-500-g08ffea331c-main/
> only the following LEAKs remain (there really should be zero leaks):
>
> - intermediate exchange: ikev2-intermediate-02-psk
> ikev2-intermediate-01-rsa-sha1
>
> +Jan 19 01:10:47.275481: leak-detective enabled
> +Jan 19 01:10:56.329908: leak: IntAuth_*_I, item size: 64
> +Jan 19 01:10:56.330122: leak: IntAuth_*_I, item size: 64
> +Jan 19 01:10:56.330192: leak: 2 * IntAuth, item size: 64
> +Jan 19 01:10:56.330345: leak detective found 6 leaks, total size 448
>
> Perhaps addressing
> https://lists.libreswan.org/pipermail/swan-dev/2020-October/004060.html
> will make this go away.
>
> - start-stop: algparse-02-fips
>
> +leak-detective enabled
> +leak: heap logger, item size: 64
> +leak: heap logger prefix, item size: 1
> +leak detective found 2 leaks, total size 65
>
> Pluto has been told to stop while addconn is still running (for
> instance when addconn hangs), which is something that should be
> handled cleanly. Ulgh.
>
> Once all these leaks are gone, I can add running refcnt.awk to the
> post mortem - unlike the leak code, it can detect Pluto leaking NSS
> objects.
>
>
>
> On Tue, 17 Nov 2020 at 17:16, Andrew Cagney <andrew.cagney at gmail.com> wrote:
> >
> > FYI,
> >
> > On Mon, 16 Nov 2020 at 09:38, Andrew Cagney <andrew.cagney at gmail.com> wrote:
> > >
> > > https://testing.libreswan.org/v4.1-178-gf93f336eb1-main
> > >
> > >       4 IntAuth_*_I
> > >       4 IntAuth
> >
> > These are from intermediate exchange.
> >
> > >       2 heap logger prefix
> > >       2 heap logger
> >
> > these are from algparse-02-fips where pluto is shutdown before it has
> > finished started up (addconn hadn't finished).
> >
> > >       1 host pair
> >
> > this, I suspect, is from adding and deleting interfaces
> >
> > >       1 ikev2_outI1 KE
> > >       1 dh
> > >
> > > and individual tests:
> > >
> > > algparse-02-fips heap logger
> > > algparse-02-fips heap logger prefix
> > > dynamic-iface-01 host pair
> > > ikev2-43-init-retransmit heap logger
> > > ikev2-43-init-retransmit heap logger prefix
> > > ikev2-43-init-retransmit ikev2_outI1 KE
> > > ikev2-43-init-retransmit dh
> > > ikev2-intermediate-01-rsa-sha1 IntAuth
> > > ikev2-intermediate-01-rsa-sha1 IntAuth
> > > ikev2-intermediate-02-psk IntAuth_*_I
> > > ikev2-intermediate-02-psk IntAuth_*_I
> > > ikev2-intermediate-02-psk IntAuth
> > > ikev2-intermediate-02-psk IntAuth_*_I
> > > ikev2-intermediate-02-psk IntAuth_*_I
> > > ikev2-intermediate-02-psk IntAuth
> > >
> > > On Fri, 30 Oct 2020 at 09:12, Andrew Cagney <andrew.cagney at gmail.com> wrote:
> > > >
> > > > This is an updated list I extracted from
> > > > https://testing.libreswan.org/v4.1-92-g55b1068bc1-main/
> > > >      42 %s/%s
> > > >      17 orphan shunt
> > > >      12 bare shunt
> > > >       4 IntAuth_*_I_A | IntAuth_*_R
> > > >       4 IntAuth_*_I
> > > >       4 2 * IntAuth
> > > >       2 responder salt
> > > >       2 initiator salt
> > > >       2 DH secret
> > > >       2 host pair
> > > >
> > > > On Wed, 28 Oct 2020 at 07:59, Andrew Cagney <andrew.cagney at gmail.com> wrote:
> > > > >
> > > > > (in case you missed the commits)
> > > > > The changes to, at the end of every test and regardless of final.sh,
> > > > > shutdown pluto then check for leaks and cores have been pushed to
> > > > > https://testing.libreswan.org/
> > > > > I've also pushed a few memory-leak fixes (one in IKEv1's aggressive
> > > > > initiator code seems to go back to at least '07).  While this makes
> > > > > some of the contents of final.sh redundant, I'll leave it for the
> > > > > moment.
> > > > >
> > > > > Below are the leaks from a fairly recent test run.   I've trimmed out
> > > > > the ones that are likely caused by crypto material being stuck in the
> > > > > limbo between helper and main threads (fix pending tm).
> > > > > This leaves what to me at least, are potential leaks on the error path:
> > > > >
> > > > > bare shunt
> > > > > host_pair
> > > > > IKEv2 fragments buffer
> > > > > ikev2_inI1outR1 KE
> > > > > initiator salt
> > > > > IntAuth
> > > > > IntAuth_*_I
> > > > > issuer ca
> > > > > libevent_malloc
> > > > > logger
> > > > > message buffer in read_packet()
> > > > > orphan shunt
> > > > > quick outR1 DH
> > > > > read_message
> > > > > responder decoding certificates
> > > > > responder salt
> > > > > %s/%s
> > > > > virtual description
> > > > > xfrmi name