Watch events are no longer sent to self
Watches no longer take a priority
async and asyncwait commands for xs_test, now we need to continue
despite blocking ops.
Print test name at end of verbose run on failure.
Use --trace-file arg to xenstored when testing
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
synchronous this makes it easier (because then we have two reasons
to block)
Instead of using return value, use explicit state member inside
struct connection.
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
Also ensure that daemon re-xmits event if they ack wrong thing
(otherwise confused clients get stuck, as we will send no more
data while awaiting ack).
Signed-off-by: Rusty Russel <rusty@rustcorp.com.au>
Signed-off-by: Christian Limpach <Christian.Limpach@cl.cam.ac.uk>
# User Rusty Russell <rusty@rustcorp.com.au>
# Node ID a92163adedcfcff0d05c965d09da747f3c8aa13e
# Parent 63ab20781afa311300f3a8e832744292014ea7f6
Remove ill-conceived concept of watches blocking reply on connection which did write/mkdir/rm/setperm etc.
This causes deadlocks in real life, and I can't see a sane way of avoiding them: it is reasonable for someone to ignore watch notifications while doing other actions, and that means that we can do other writes. These writes can block pending other watchers; if one of these is the process blocked awaiting our ack, we deadlock.
diff -r 63ab20781afa -r a92163adedcf tools/xenstore/xenstored_core.c
active at startup; this works by adding a module line into the grub
configuration file and placing the binary policy generated by the
policy tool into the boot directory; This assumes that a maximum of
one module line is used for the initrd in the grub configuration
file - Question: do users pass more than one module to the kernel?
- enables the policy hypervisor call on x86/64
- some function prototypes moved to include files
- moves the version number in the java tool up to the current
version (a better way of doing this will be submitted soon)
Signed-off-by: Stefan Berger <stefanb@us.ibm.com>
Signed-off-by: Reiner Sailer <sailer@us.ibm.com>
Even though we accounted for lost ticks, the accounting was not precise.
Specifically, we didn't account for the delay in calling the ac_timer
handler.
Signed-off-by: Eddie Dong <eddie.dong@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
Missed out one instance where we need to check if we're running on
1:1 page tables with guest paging disabled.
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
This enables cdrom ejection (both file based and physical i.e. /dev/cdrom)
Signed-off-by: Edwin Zhai <edwin.zhai@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
With this patch, we should not bee seeing any vmexits for debug port
accesses.
Signed-off-by: Edwin Zhai <edwin.zhai@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
This removes odd control characters displayed in the qemu monitor with vnc
Signed-off-by: Edwin Zhain <edwin.zhai@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
The original gdb server patch will cause the new vmx guest break the=20
old one. This patch make the arch_set_guest_info modify the correct
vmcs and also skip modifying during creating.
Signed-off-by: Xiaofeng Ling <xiaofeng.ling@intel.com>^
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
Without this, we may be saving the IDT info from the wrong cpu into the
vmcs.
This makes domU work in the presence of VMX domains on SMP/HT enabled
machines.
Signed-off-by: Chengyuan Li <chengyuan.li@intel.com>
Signed-off-by: Xiaofeng Ling <xiaofeng.ling@intel.com>=20
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
Right now, we have an exposure between the time the MSR is written
and used by an instruction such as syscall. If there is a context
switch and we do vmx_do_restore_msrs(), everything goes fine. But
if we don't, then we execute the syscall with the wrong MSR.
Signed-off-by: Yunhong Jiang <yunhong.jiang@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
Based on an initial patch from Don Fry at IBM.
Still TODO:
1. Testing
2. NTP synchronisation
3. Fix wallclock interface a bit
4. Support for platform timers other than PIT (e.g., HPET, IBM Cyclone)
5. Scale 64-bit TSC diffs instead of 32-bit, just for sanity
6. Error-correcting scale factor is still slightly wrong
6. More testing
Signed-off-by: Keir Fraser <keir@xensource.com>
Pull out PAL_VM_SUMMARY, reimplement later
Turn off ar.unat code in fast paths by default for now
Fix vcpu_fc
Signed-off by: Dan Magenheimer <dan.magenheimer@hp.com>
