Re: CURRENT r255426: x11/nvidia-driver: nvidia_subr.c:835:46: error: too few arguments to function call, expected 10, have 9 (VM_PROT_READ | VM_PROT_WRITE), 0);

On Tue, 2013-09-10 at 00:42 +0200, O. Hartmann wrote:
> On Tue, 10 Sep 2013 00:24:05 +0300
> Ivan Klymenko <fidaj_at_ukr.net> wrote:
> 
> > В Mon, 9 Sep 2013 23:17:13 +0200
> > "O. Hartmann" <ohartman_at_zedat.fu-berlin.de> пишет:
> > 
> > > On Tue, 10 Sep 2013 00:02:37 +0300
> > > Ivan Klymenko <fidaj_at_ukr.net> wrote:
> > > 
> > > > В Mon, 9 Sep 2013 22:37:48 +0200
> > > > "O. Hartmann" <ohartman_at_zedat.fu-berlin.de> пишет:
> > > > >  /usr/ports/emulators/virtualbox-ose-kmod/work/VirtualBox-4.2.18/out/freebsd.amd64/release/bin/src/vboxdrv/r0drv/freebsd/alloc-r0drv-freebsd.c:83:76:
> > > > > error: too few arguments to function call, expected 10, have 9
> > > > > cbAllocated, TRUE, VM_PROT_ALL, VM_PROT_ALL, 0); ^
> > > > > _at_/vm/vm_map.h:368:1: note: 'vm_map_find' declared here int
> > > > > vm_map_find(vm_map_t, vm_object_t, vm_ooffset_t, vm_offset_t *,
> > > > > vm_size_t,
> > > > 
> > > > Try the following changes:
> > > >         int rc = vm_map_find(kernel_map, pVmObject, 0, &Addr,
> > > > ---                          cbAllocated, TRUE, VM_PROT_ALL,
> > > > VM_PROT_ALL, 0); +++                          cbAllocated, 0,
> > > > VMFS_OPTIMAL_SPACE, VM_PROT_ALL, VM_PROT_ALL, 0);
> > > > 
> > > > for Nvidia driver need to make similar changes...
> > > The man page of vm_map_find() says the last parameter is of int,
> > > named int cow. On the hurry, I didn't find any explanation of that
> > > parameter. Setting it to "0" (zero) also in the nvidia_subr.c makes
> > > the driver compile again.
> > > 
> > > But simply filling in a int zero is a bit strange without knowing
> > > what to do, isn't it?
> > > 
> > > Thanks anyway,
> > > 
> > > Oliver
> > 
> > This commit you to find an appropriate example ...
> > http://docs.freebsd.org/cgi/getmsg.cgi?fetch=291456+0+current/svn-src-head
> > And in this case, the driver Nvidia really need to specify 0 in place
> > of the sixth argument ...
> > but I'm not sure at 100% :)
> 
> I have a patch attached for the x11/nvidia-driver Makefile. It doesn't
> work and I loose hairs due to not knowing why.
> 
> When issuing the following command sequence on the console:
> 
> cd /usr/ports/x11/nvidia-driver
> make clean extract
> sed -i -e '/(VM_PROT_READ | VM_PROT_WRITE), 0);$/s/0);$/0, 0);/g' \
> work/NVIDIA-FreeBSD-x86_64-325.15/src/nvidia_subr.c
> 
> 
> and check then line 835 (the corrupt one) in file 
> work/NVIDIA-FreeBSD-x86_64-325.15/src/nvidia_subr.c,
> 
> I see this BEFORE:
> 
>     status = vm_map_find(kernel_map, at->object, (i * PAGE_SIZE),
>             &virtual_address, size, VMFS_ANY_SPACE,
>             (VM_PROT_READ | VM_PROT_WRITE),
> ===(835)>>            (VM_PROT_READ | VM_PROT_WRITE), 0);
> 
> 
> and I see this AFTER the sed'ed replacement:
> 
>     status = vm_map_find(kernel_map, at->object, (i * PAGE_SIZE),
>             &virtual_address, size, VMFS_ANY_SPACE,
>             (VM_PROT_READ | VM_PROT_WRITE),
> ===(835)>>            (VM_PROT_READ | VM_PROT_WRITE), 0, 0);
> 
> BUT: Using the patch (see attached, please apply to
> x11/nvidia-driver/Makefile) gives me an error in x11/nvidia-driver:
> 
> make clean patch
> 
> ===>  Cleaning for nvidia-driver-325.15
> ===>  License NVIDIA accepted by the user
> ===>  Found saved configuration for nvidia-driver-325.08_1
> ===>   nvidia-driver-325.15 depends on file: /usr/local/sbin/pkg - found
> ===> Fetching all distfiles required by nvidia-driver-325.15 for
> building
> ===>  Extracting for nvidia-driver-325.15
> => SHA256 Checksum OK for NVIDIA-FreeBSD-x86_64-325.15.tar.gz.
> ===>  Patching for nvidia-driver-325.15
> sed: 1: "/(VM_PROT_READ | VM_PRO ...": invalid command code 0
> *** Error code 1
> 
> The Shell/make should be able to substitute within a single-quotet
> command to sed, but the error message reports differently.
> 
> Do not be confused about the driver revision I use. I tried the
> official one (319.XX) as well and it is not working the very same way
> and the patch won't either.
> 
> Oliver

I don't know about the sed error, but I think the change you're trying
to make is wrong.  The new '0' parameter to vm_map_find() isn't added at
the end, it goes after the size parameter.

-- Ian