[uml-devel] 2.6-specific UML patches (modules, modversions, various other stuff)

[uml-devel] 2.6-specific UML patches (modules, modversions, various other stuff)

[BlaisorBlade]

Hi! I'm happy to hear again from you.

I don't remember if you answered when I posted this URL:
http://web.tiscali.it/no-redirect-tiscali/blaisorblade/

There you will find a number of patches for UML 2.6. Some are mine and some 
come from Ingo and M A Young; I've included everything Ingo posted minus the 
speedup - breakage for 2.4 host. That is, MM_COPY_SEGMENTS is still called by 
UML but if the host is a 2.6 that is a no-op.

Only use the v8! In the Readme there is a careful description of each patch - 
ask for more comments if you need.

They are tested enough (at least for building UML; the resulting patch should 
work even on i386, and at least it compiles). Maybe you already have in your 
tree some of the patches; they apply onto vanilla linux kernel. The -test9 
patch has been split: the change for tty_init has been removed, and the 
change to mm/memory.c has been replaced. You can find the two patches to move 
out here, as A-03* and A-04*:

http://web.tiscali.it/no-redirect-tiscali/blaisorblade/linux/archives/UML/v6/A-Orig/

The replacements are in the B directory. About the A-04 patch, and this hunk 
in particular(it goes to mm/memory.c):

@@ -700,19 +719,8 @@

                vma = find_extend_vma(mm, start);

-#ifdef FIXADDR_USER_START
-               if (!vma &&
-                   start >= FIXADDR_USER_START && start < FIXADDR_USER_END) {
-                       static struct vm_area_struct fixmap_vma = {
-                               /* Catch users - if there are any valid
-                                  ones, we can make this be "&init_mm" or
-                                  something.  */
-                               .vm_mm = NULL,
-                               .vm_start = FIXADDR_USER_START,
-                               .vm_end = FIXADDR_USER_END,
-                               .vm_page_prot = PAGE_READONLY,
-                               .vm_flags = VM_READ | VM_EXEC,
-                       };
+#ifdef FIXADDR_START
+               if (!vma && start >= FIXADDR_START && start < FIXADDR_TOP) {
                        unsigned long pg = start & PAGE_MASK;
                        pgd_t *pgd;
                        pmd_t *pmd;
Did you have any reason for replacing FIXADDR_USER_START with FIXADDR_START?
I've looked in older patches and saw that FIXADDR_USER_START didn't exist, so 
I thought that you simply didn't see the change happened in mainline; even 
because I've maybe understood what code does, and there is no need to map 
anything in a process's address space except for the VSYSCALL page. Am I 
right?

I must admit that with my replacement for that patch, I've got one 
non-repeatable segfault, in the early boot process. I've never had it any 
more, so maybe it's elsewhere.

For the makefile patches, they haven't still been accepted by Kbuild 
developers :-(. But everything is open (kbuild mail list is not that active, 
actually...)

Bye
-- 
cat <<EOSIGN
Paolo Giarrusso, aka Blaisorblade
Linux Kernel 2.4.23/2.6.0 on an i686; Linux registered user n. 292729
EOSIGN



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: 2.6-specific UML patches (modules, modversions, various other stuff)

[Jeff Dike]

blaisorblade_spam@yahoo.it said:
> There you will find a number of patches for UML 2.6. Some are mine and
> some  come from Ingo and M A Young; I've included everything Ingo
> posted minus the  speedup - breakage for 2.4 host. That is,
> MM_COPY_SEGMENTS is still called by  UML but if the host is a 2.6 that
> is a no-op. 

There's a lot of old stuff in that patch.  It would be more useful to me if
it were against a recent UML patch, like the 2.6.0 patch that I'm working
on at the moment.  That has all the recent 2.4 changes pulled in as well as
updating to 2.6.0.  A patch containing the stuff that you have that I don't
would be useful.

Some more specific things -
	I'm not going to update the skas3 patch again, but I'm glad people are
cleaning it up and updating it.
	I don't like the UML_USER_OBJS thing - feel free to work on it, but
I'm not going to merge it unless I'm desperate.  I continue to think that the
right way to go is to move all the userspace stuff under os-Linux, then fiddle
the build rules there.

				Jeff



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: 2.6-specific UML patches (modules, modversions, various other stuff)

[BlaisorBlade]

Alle 20:09, martedì 6 gennaio 2004, Jeff Dike ha scritto:
> blaisorblade_spam@yahoo.it said:
> > There you will find a number of patches for UML 2.6. Some are mine and
> > some  come from Ingo and M A Young; I've included everything Ingo
> > posted minus the  speedup - breakage for 2.4 host. That is,
> > MM_COPY_SEGMENTS is still called by  UML but if the host is a 2.6 that
> > is a no-op.
>
> There's a lot of old stuff in that patch.  It would be more useful to me if
> it were against a recent UML patch, like the 2.6.0 patch that I'm working
> on at the moment.  That has all the recent 2.4 changes pulled in as well as
> updating to 2.6.0.  A patch containing the stuff that you have that I don't
> would be useful.

I've not got your patch, as I'm not a wizard :-), but the split patches should 
be ok(also you can choose only what you like easily). Just don't take neither 
the combo nor the A-* patches (they just contain the stuff from the -test9 
patch). And remove the changes I pointed out as A-03 and A-04 in the original 
message.

Did you read my question about the change inside mm/memory.c you did?

> Some more specific things -
> 	I'm not going to update the skas3 patch again, but I'm glad people are
> cleaning it up and updating it.
Ok, that is inside E* and F* patches. I've worked on split-up because anyway 
those bit couldn't go in mainline (while I hope that Andrew Morton will soon 
accept the UML patch...)

> 	I don't like the UML_USER_OBJS thing - feel free to work on it, but
> I'm not going to merge it unless I'm desperate.  I continue to think that
> the right way to go is to move all the userspace stuff under os-Linux, then
> fiddle the build rules there.

Just note that without that CONFIG_MODVERSIONS maybe doesn't work (kbuilds 
postprocess the files with modversions, see scripts/Makefile.lib, but not 
without that rule). Actually, I've fixed the link scripts too before 
obtaining success with modversions (I didn't test before), and maybe so that 
is not needed. Also, nothing need that patch apart H-03 - another kbuild 
cleanup, much less intrusive; most of it makes still sense, but it will have 
too much rejects; I'll later post a more up-to-date version of it. H-04 (the 
Pcap fix) should apply anyhow.

About your idea: that is meaningful, but hostfs and other stuff can be built 
as modules. Then is what you get in that way:

1) hostfs_user goes into os-Linux
2) hostfs_kern remains inside fs/hostfs
3) kbuild will refuse to link together hostfs_user and hostfs_kern inside a 
module.
However, this can be solved by making 2 separate modules... (and playing with 
exports between modules if needed).

Also, that doesn't remove the need for fiddling with main kbuild 
infrastructure, since otherwise:
1) either you duplicate a lot of stuff from main Makefiles for .c -> .o rule
2) or you change the c_flags variables from arch/um/os-Linux makefile, 
fiddling with, but that is fragile as that is an undocumented interface.

Bye
-- 
cat <<EOSIGN
Paolo Giarrusso, aka Blaisorblade
Linux Kernel 2.4.23/2.6.0 on an i686; Linux registered user n. 292729
EOSIGN



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id\x1278&alloc_id371&opÌk
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] 2.6-specific UML patches (modules, modversions, various other stuff)

[Ingo Molnar]

* BlaisorBlade <blaisorblade_spam@yahoo.it> wrote:

> There you will find a number of patches for UML 2.6. Some are mine and
> some come from Ingo and M A Young; I've included everything Ingo
> posted minus the speedup - breakage for 2.4 host. That is,
> MM_COPY_SEGMENTS is still called by UML but if the host is a 2.6 that
> is a no-op.

i'd suggest to skip my speedup altogether, skas4 ought to make it
obsolete.

i think it still makes sense to keep host-skas3 working in the 2.6 combo
patch until Jeff releases skas4 - the positive performance impact from
the host-skas concept is significant and the difference between skas3
and skas4 should not affect performance much, it's 'just' a (much)
cleaner interface.

	Ingo


-------------------------------------------------------
This SF.net email is sponsored by: Perforce Software.
Perforce is the Fast Software Configuration Management System offering
advanced branching capabilities and atomic changes on 50+ platforms.
Free Eval! http://www.perforce.com/perforce/loadprog.html
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel