[Xen-devel] guest debug support patches

[Kip Macy <kip.macy@xxxxxxxxx>]

These two patches should really be one patch. However, when I try:

bk export -tpatch -r1.1294.1,1.1307,1.1308,1.1309,1.1310,1.1311

The output includes a large number of files that I haven't touched. My
guess is that there is a bug in the graph re-labelling in bk.

Thus patch1 is the output of:
bk export -tpatch -r1.1294.1.1

And patch2 is the output of:
 bk export -tpatch -r1.1307,1.1308,1.1309,1.1310,1.1311

These patches have three real changes:
1) xen/arch/x86/traps.c in do_int3 and do_debug if the trap was hit in
kernel-mode on domain other than dom0 the domain is paused

2) builddomain is re-factored into setdomaininfo, the parts of what
was arch_final_setup_guest that should only be set once are skipped by
means of a new exec_domain flag EDF_DONEINIT

3) xc_ptrace.c implements the ptrace interface via dom0 ops 

known limitations:
- Although the hypervisor-side code should work on x86_64, the
user-side xc_ptrace is currently strictly x86_32, fixing this is a
SMOP

- xc_ptrace assumes the guest is UP - I intend to fix this right after
I get 5.3 booting SMP as a guest - I would wait until this was fixed
were it not for the fact that the resolves can get annoying

- the patch doesn't include the gdbserver side support, I'm not sure
how best to handle this as most GDB maintainers move at a glacial pace
when it comes to incorporating outside patches
 
Comments and suggestions are welcome.

signed off by Kip Macy kmacy at fsmware.com

Attachment: patch1
Description: Text document

Attachment: patch2
Description: Text document