Date: Sat, 13 Jul 2013 14:18:26 +0300 From: Mikolaj Golub <trociny@FreeBSD.org> To: Craig Rodrigues <rodrigc@FreeBSD.org> Cc: svn-src-head@freebsd.org, svn-src-all@freebsd.org, Adrian Chadd <adrian@freebsd.org>, src-committers@freebsd.org, Marko Zec <zec@fer.hr> Subject: Re: svn commit: r253255 - head/sys/netgraph/bluetooth/drivers/ubt Message-ID: <20130713111825.GE2757@gmail.com> In-Reply-To: <CAG=rPVe_F03LTt1K77p1BSw0hojMm-Ms-oZdhaFgo842AxkmhQ@mail.gmail.com> References: <201307120803.r6C83ACp016523@svn.freebsd.org> <CAJ-Vmonrmgr3HJRzQtVmWp=8SkpSeWNVeBZgOnSnE72NNp=3DQ@mail.gmail.com> <CAG=rPVcxvWq-MoXADiv%2BgLY-d3hhv1GTMO1R14bhLRm9UB9DqA@mail.gmail.com> <20130712202710.GB2757@gmail.com> <CAJ-Vmomv1nEEq9HV9GreQg6KJGos6ffDhFcVUsUuzfCL6h-YQw@mail.gmail.com> <CAJ-Vmokm29XXbQr5uTwEbGGDYo-8WQNsp60b%2B7Khkteyh%2BO%2B6A@mail.gmail.com> <CAJ-VmomidFufKmGUb0YP2pHWu%2BpF2Os3Fo4JWzkMJDKLqP_ing@mail.gmail.com> <20130713073118.GC2757@gmail.com> <CAG=rPVe_F03LTt1K77p1BSw0hojMm-Ms-oZdhaFgo842AxkmhQ@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Jul 13, 2013 at 03:31:34AM -0700, Craig Rodrigues wrote:
>
> I experimented with Marko's patch, and found that the kernel crashed
> inside the code which initializes PF.
....
> /usr/home/rodrigc/freebsd/9/sys/modules/pf/../../contrib/pf/net/pf_ioctl.c:295
> 295 {
> (kgdb) l
> 290
> 291 mtx_destroy(&pf_task_mtx);
> 292 }
> 293 void
> 294 init_zone_var(void)
> 295 {
> 296 V_pf_src_tree_pl = V_pf_rule_pl = NULL;
> 297 V_pf_state_pl = V_pf_state_key_pl = V_pf_state_item_pl =
> NULL;
> 298 V_pf_altq_pl = V_pf_pooladdr_pl = NULL;
> 299 V_pf_frent_pl = V_pf_frag_pl = V_pf_cache_pl = V_pf_cent_pl
> = NULL;
>
> In init_zone_var(), curthread->td_vnet was NULL.
I wonder if it fails on current too? It looks this code is not there
any more. I will try.
> So, I don't think it is safe to remove the vnet stuff from
> kldload/kldunload just yet,
> without fixing some code in some kernel modules like PF.
>
If you have plans to merge the changes to stable then yes, I think it
is safer to commit the patch without kldload/kldunload part and commit
the last separately after testing (and fixing affected modules if
there are).
> The other parts of the patch look OK.
So it works for you with your commit reverted and without Adrians'
netgraph patch?
--
Mikolaj Golub
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130713111825.GE2757>