Re: [ovs-dev] memory leak in recirc_state?

> On Jan 6, 2016, at 1:21 PM, Ben Pfaff wrote: > > On Wed, Jan 06, 2016 at 01:14:23PM -0800, Jarno Rajahalme wrote: >> >>> On Jan 5, 2016, at 8:24 PM, Ben Pfaff wrote: >>> >>> Thanks. That was my experience also. >>> >>> I applied this to master and branch-2.5.

Re: [ovs-dev] memory leak in recirc_state?

On Wed, Jan 06, 2016 at 01:14:23PM -0800, Jarno Rajahalme wrote: > > > On Jan 5, 2016, at 8:24 PM, Ben Pfaff wrote: > > > > Thanks. That was my experience also. > > > > I applied this to master and branch-2.5. > > Thanks for fixing this. I reviewed the patch and it seems

Re: [ovs-dev] memory leak in recirc_state?

> On Jan 5, 2016, at 8:24 PM, Ben Pfaff wrote: > > Thanks. That was my experience also. > > I applied this to master and branch-2.5. Thanks for fixing this. I reviewed the patch and it seems correct to me as well. > I think that branch-2.4 has > the same bug but the backport

Re: [ovs-dev] memory leak in recirc_state?

Hi Ben, These two tests generate the leak: mpls_xlate 381: MPLS xlate action ofproto-dpif 852: ofproto-dpif - MPLS handling ==65139==by 0x4E1C83: xmemdup (util.c:134) ==65139==by 0x431044: recirc_state_clone (ofproto-dpif-rid.c:221) ==65139==by 0x431044: recirc_alloc_id__

[ovs-dev] memory leak in recirc_state?

Looking at the recirculation code in ofproto-dpif-rid.c, I see code to copy the stack and the actions from one recirc_state to another in recirc_state_clone(), but I can't find any code that ever frees them. recirc_run(), for example, just does a plain (postponed) "free" of the recirc_id_node

Re: [ovs-dev] memory leak in recirc_state?

I think that recirc_run needs to be modified so that On Tue, Jan 05, 2016 at 06:40:23PM +, ChengChun Tu wrote: > Hi Ben, > > Yes, Valgrind testcase 381 reports leak and generates the call stack below: > I tried to debug it for a while but not able to understand it. Thanks, what's the name

Re: [ovs-dev] memory leak in recirc_state?

Hi Ben, Yes, Valgrind testcase 381 reports leak and generates the call stack below: I tried to debug it for a while but not able to understand it. fun:malloc fun:xmalloc fun:ofpbuf_new fun:ofpbuf_new_with_headroom fun:ofpbuf_clone_data_with_headroom fun:ofpbuf_clone_with_headroom

Re: [ovs-dev] memory leak in recirc_state?

Thanks. That was my experience also. I applied this to master and branch-2.5. I think that branch-2.4 has the same bug but the backport is not trivial and I do not know whether it is worthwhile. On Tue, Jan 05, 2016 at 05:22:24PM -0800, William Tu wrote: > Hi Ben, > > The patch works OK. It

Re: [ovs-dev] memory leak in recirc_state?

Thanks. Would you mind testing this proposed fix? http://openvswitch.org/pipermail/dev/2016-January/064070.html On Tue, Jan 05, 2016 at 11:32:48AM -0800, William Tu wrote: > Hi Ben, > > These two tests generate the leak: > mpls_xlate > 381: MPLS xlate action > ofproto-dpif >

Re: [ovs-dev] memory leak in recirc_state?

Hi Ben, Sure, I will test this fix. Regards, William On Tue, Jan 5, 2016 at 4:55 PM, Ben Pfaff wrote: > Thanks. Would you mind testing this proposed fix? > http://openvswitch.org/pipermail/dev/2016-January/064070.html > > On Tue, Jan 05, 2016 at 11:32:48AM -0800,

Re: [ovs-dev] memory leak in recirc_state?

Hi Ben, The patch works OK. It passes "make check" and "make check-valgrind" without reporting memory leaks. Thank you William On Tue, Jan 5, 2016 at 5:02 PM, William Tu wrote: > Hi Ben, > > Sure, I will test this fix. > > Regards, > William > > On Tue, Jan 5, 2016 at 4:55