Fixes to make the vnet module compile under 2.6.10.
Thanks to Brian Bruce for spotting the problem and
suggesting fixes.
Signed-off-by: Mike Wray <mike.wray@hp.com>
Subject: [Xen-devel] [PATCH] Fix initrd support for Linux 2.4/2.6
The following patch sets the default root device to be the null device.
This is the behavior Linux is launched with no root device specified in
the command line and without a device provided by the boot loader.
Previously, the default root device was set to be /dev/ram0. This broke
initrd support. This change was made in 2.6 to because of a bug
reported on September 14th with dynamic ipconfig. I've recreated his
problem and looked through the code and the source of the problem was
that his kernel command line was incorrect. He was missing the BOOTP
field in the config line requiring that there be an actual root device.
This worked with 2.4 but only because 2.4 was also broken. His patch
just introduced that bug into 2.6.
I've tested this against the 2.6 and 2.4 trees in a recent copy of
xen-unstable. I've tested initrd support and ipconfig support
(including dynamic ip support). They both now work as expected.
If anyone needs an initrd for testing let me know.
Regards,
--
Anthony Liguori
Linux Technology Center (LTC) - IBM Austin
E-mail: aliguori@us.ibm.com
Phone: (512) 838-1208
Signed-off-by: Anthony Liguori
Signed-off-by: ian@xensource.com
Some functions aren't static and could be (damn C language!).
I tried turning on -Wmissing-prototypes: unfortunately gives warnings
for functions used in asm, which means introducing gratuitous prototypes
for them. Not sure it's worth it.
1) keyhandler.c: keypress_softirq() and do_task_queues() can be static.
2) physdev.c: pcidev_dom0_hidden() can be static.
3) resource.c/resource.h: check_region is deprecated (racy): remove.
4) sched_bvt.c: lots of things can be static.
5) pci/compat.c: not required for Xen.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (authored)
Signed-off-by: ian.pratt@cl.cam.ac.uk
This very simple patch is really just cosmetic. It "fixes" xend so that
it doesn't report restart as: "WARNING (XendDomainInfo:1038) Unknown
config field restart" in the xend.log. I'm not sure who should sign off
on this, but the patch is attached.
From: "B.G. Bruce" <bgb@nt-nv.com>
Signed-off-by: ian@xensource.com
As pointed out on the xen-devel list, the vm86 call was not very robust in
Xen.
It turned out to be due to a missing check in the pagefault handler,
determining whether the pagefault came from userspace or from the kernel.
Now the syscall-crash tool doesn't cause any "Oops" on the vm86 calls anymore.
From: Stephan Diestelhorst <sd386@cam.ac.uk>
Signed-off-by: ian.pratt@cl.cam.ac.uk
Bug fix to the scheduler for writable page tables.
Don't flush the writable page tables inside __enter_scheduler()
until after we've committed to switching to another domain.
Here's the problem: if enough debugging messages are enabled (for
example, shadow mode and/or writable page tables messages), then we
can get into a situation where we *always* visit
__enter_scheduler() after every page fault and/or hypercall. If we
flush the writable page tables every time we take a page fault,
we'll never end up actually getting into the domain with write
permissions enabled to its page table.
Signed-off-by: michael.fetterman@cl.cam.ac.uk
Name: pic-lib.patch
Description: support PIC code generation
Libraries have 2 methods of being compiled; a .so must contain non-pic
objects, while the .a can contain standard objects.
Signed-off-by: Adam Heath <doogie@brainfood.com>
Signed-off-by: ian.pratt@cl.cam.ac.uk
Name: tools-misc-TARGETS.patch
Description: Split compile/link stages
distcc behaves much better when you compile and link in 2 separate steps.
Plus, this is a more standard way of doing things.
Signed-off-by: Adam Heath <doogie@brainfood.com>
Signed-off-by: ian.pratt@cl.cam.ac.uk
Small hack to make CROMMULTISESSION ioctl return success rather than enosys, like we did on Xen 1.2.
From: Anthony Liguori aliguori@us.ibm.com
Signed-off-by: ian.pratt@cl.cam.ac.uk
Fix the device number calculation for /dev/hd* device names;
ide device numbering works differently to scsi
Signed-off-by: Jody Belka <knew-xen@pimb.org>
Signed-off-by: ian.pratt@cl.cam.ac.uk
No longer override target arch with ARCH but instead use TARGET_ARCH.
Xen now requires specification of TARGET_ARCH alone, not TARGET_ARCH
and TARGET_SUBARCH. Xen derives correct internal ARCH/SUBARCH components
automatically.
