The scandalous status of linux-flashplugin9

[Juergen Lock]

[adding -emulation to Cc, just in case...]

On Wed, Oct 10, 2007 at 09:40:55PM +0200, I wrote:
> In article <20071008192953.677522d5 at deskjail> you write:
> >Quoting John Reynolds <johnjen at reynoldsnet.org> (Mon, 8 Oct 2007 08:10:11 -0700):
> >
> >> 
> >> [ On Monday, October 8, Scot Hetzel wrote: ]
> >> > The port is not broken as the flash9 port is not compiled, it just
> >> > installs the linux flash9 binary.  What is broken is the linux
> >> > emulation on FreeBSD < 7. Work is underway to improve the linux
> >> > emulation in -CURRENT.
> >> > 
> >> > I agree the port should be marked broken for OSVERSION < 7000xx, and
> >> > compat.linux.osrelease = 2.4.2, as the flash9 plugin may require
> >> > 2.6.16 linux emulation.
> >> 
> >> Is there a handbook or wiki entry which would point one to the procedure for
> >> "properly" moving from 2.4.2 to 2.6.16 emulation under -current? I believe I
> >> saw some posts that 2.6.x would not be default for 7.0-RELEASE for whatever
> >> technical reason.....
> >
> >% grep linux /etc/sysctl.conf
> >compat.linux.osrelease=2.6.16
> >
> >Be warned, you may run into bugs (2.6 emulation has not as much
> >widespread testing as 2.4). If you stumble upon problems have a look at
> >the archive of emulation@ and search for your problem there. If you can
> >not find it, tell emulation@ about your problem. Be prepared to do some
> >guided debugging.
> >
> >> Can anybody report success on -current with 2.6.16 emulation with the flash9
> >> plugin? 
> >
> >No, there are bugs we didn't manage to track down yet (partly because of
> >lack of time, partly because of the closed source nature of flash9,).
> >If someone wants to help to track this down: install dtrace (I don't
> >know where install instructions are; it's back from hibernation just
> >recently), run flash9 and try to get a backtrace of a crash with
> >dtrace. If someone manages to do this, post the backtrace to emulation at .
> >
> >Bye,
> >Alexander.
> 
> There was a report on -current a while ago that a preliminary patch
> to fix the non-threadsafeness of mmap(2) MAP_FIXED also makes flash9 work,
> 	http://lists.freebsd.org/pipermail/freebsd-current/2007-August/075968.html
> a fix for that seems to have been committed (kib 2007-08-20) and also
> mfc'd to RELENG_6 by now (kib 2007-09-09), so, can anyone running either
> verify that flash9 still crashes for them?

Ok I now have a sucess story for RELENG_7 using linux-base-fc4
(hi wallshot! :) - he couldnt get it to work with linux-base-f7), and
I myself (he tested too) couldnt get it working by merging the RELENG_6
commit to 6.2, or by testing a RELENG_6 kernel and linux.ko on 6.2 userland.
So it looks like you need 7, and there's still work to be done for the
linux 2.6 emulation there also.

 Oh, here's the patch I used on 6.2:
	http://people.freebsd.org/~nox/mmap-6.2.patch
meged from this commit:

kib         2007-09-09 04:41:24 UTC

  FreeBSD src repository

  Modified files:        (Branch: RELENG_6)
    sys/vm               vm_map.c vm_map.h vm_mmap.c 
  Log:
  MFC
  rev. 1.387 of src/sys/vm/vm_map.c
  rev. 1.120 of src/sys/vm/vm_map.h
  rev. 1.213 of src/sys/vm/vm_mmap.c
  
  Do not drop vm_map lock between doing vm_map_remove() and vm_map_insert().
  For this, introduce vm_map_fixed() that does that for MAP_FIXED case.
  
  Tested by:      Marc G. Fournier <scrappy hub org>
  
  Revision   Changes    Path
  1.366.2.6  +35 -16    src/sys/vm/vm_map.c
  1.117.2.2  +1 -0      src/sys/vm/vm_map.h
  1.200.2.4  +4 -2      src/sys/vm/vm_mmap.c

 HTH,
	Juergen