Date: Fri, 26 Oct 2007 00:53:13 +0400 From: pluknet <pluknet@gmail.com> To: "John Baldwin" <jhb@freebsd.org> Cc: anholt@freebsd.org, freebsd-current@freebsd.org Subject: Re: agp.ko panic: vm_fault: fault on nofault entry, addr: deadc000 Message-ID: <a31046fc0710251353n746d469dg963c9a8bbb616d4d@mail.gmail.com> In-Reply-To: <200710251434.18764.jhb@freebsd.org> References: <a31046fc0710201930w75e2d46aicb41599f3ec79752@mail.gmail.com> <200710251434.18764.jhb@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 25/10/2007, John Baldwin <jhb@freebsd.org> wrote: > On Saturday 20 October 2007 10:30:31 pm pluknet wrote: > > Hello all. > > > > I could produce panic while kldunload'ing agp.ko. > > > > It's on 7.0-CURRENT from Oct 11 (just before RELENG_7), i386, UP, > > Intel 82855GME (855GME GMCH) SVGA controller. > > Some debugging below: > > > > panic: vm_fault: fault on nofault entry, addr: deadc000 > > KDB: enter: panic > > This is caused by the agp_i810() driver calling agp_generic_detach() first > which frees the memory resource for the aperture. Other AGP drivers do the > same thing. I think the proper fix is to split agp_generic_detach() into two > pieces. The first part would do a destroy_dev() of the /dev node and the > various agp_foo_detach() methods would call that first. The rest > agp_generic_detach() would be called at the end of the agp_foo_detach() > methods. Yes, I forgot to note that there was a message about memory leakage on panic. > > A possible patch (untested) is at > http://www.FreeBSD.org/~jhb/patches/agp_detach.patch and included below: Thank you. This patch fixes my issue. Now all three modules (drm, agp, i915) unloaded without a panic. > > -- > John Baldwin > wbr, pluknet
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?a31046fc0710251353n746d469dg963c9a8bbb616d4d>