* Add traces for credit reset and scheduling a tasklet
* Remove tsc for traces which probably don't need them
* Print domain info in the debug dump
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
The previous code had a bug where, if a second vcpu woke up and ran
runq_tickle before the first vcpu had actually started running on a
tickled processor, the code would choose the same cpu to tickle, and
the result would be a "missed tickle" -- only one of the vcpus would
run, despite having idle processors.
This change:
* Keeps a mask of idle cpus
* Keeps a mask of cpus which have been tickled, but not entered
schedule yet.
The new tickle code first looks at the set of idle-but-not-tickled
cpus; if it's not empty, it tickles one.
If that doesn't work, it looks at the set of not-idle-but-not-tickled
cpus, finds the one with the lowest credit; if that's lower than the
waking vcpu, it tickles that one.
If any cpu is tickled, its bit in the tickled mask is set, and cleared
when schedule() is called.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Sometimes we would rather connect to an HVM's serial console than it's
VGA display. In this case we cannot see or interact with the ethernet
booting process. This is a bit annoying so let's just enable it.
Signed-off-by: Gianni Tedesco <gianni.tedesco@citrix.com>
When using cpupools don't pin vcpus to numa nodes as this might
conflict with the cpupool definition.
numa placement should be handled by cpupool configuration instead.
Signed-off-by: Juergen Gross <juergen.gross@ts.fujitsu.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
In the default "QemuDeviceModelRecord"-style HVM tail of the migration
protocol the size of the qemu save record is unknown at the receiver
and therefore it must read until EOF. This is not compatible with the
xl migration protocol which contains a post-migration handshake and
therefore cannot close the socket on the sending end.
What is required is an explicit length field for the save record,
which the "RemusDeviceModelState"-style HVM tail includes.
Rather than overload the "RemusDeviceModelState" name for a non-Remus
use case (on off chance that they need to diverge for some reason in
the future) introduce a third style called "DeviceModelRecord0002"
which is identical to current "RemusDeviceModelState"-style.
Hopefully the inclusion of a number here will allow easier extension
in the future without needing to come up with increasingly less
helpful names!
Also propagate errors from xc_domain_save and
libxl__domain_suspend_common to callers.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
According to the AMD64 ABI, the 128-byte area below %rsp is reserved
for scratch space and should not be used by signal, interrupt, or
exception handlers. However, Mini-OS does not adhere to this
interface. As a result, the -mno-red-zone flag must be provided to gcc
in order to force code to be compiled without this assumption. In the
case of the c-stubdom or the ioemu stubdom, Makefile configuration and
flags are inherited from Xen, which includes this flag. However, in
the case of the ocaml stubdom, since an independent ocaml version is
pulled and compiled, with its own make system in place, this flag is
not present.
As a default optimisation (no optimisation flags specified), gcc
generates function header code that uses mov instead of push (to save
a few instructions). However, with the 128-byte scratch space
assumption in place, registers are moved onto the stack *before* %rsp
is updated. This results in cases where when a function is being
called, *after* the first mov <foo>, <offset>(%rsp) but *before* the
%rsp update, sub <bar>, %rsp, if an interrupt is fired, then the stack
will be clobbered (as Mini-OS uses the application stack for its
stack, pushing registers onto the stack on entry and popping them off
on exit).
This patch fixes this issue by passing the -mno-red-zone flag to the
ocaml configure script in the stubdom Makefile.
Signed-off-by: Patrick Colp <pjcolp@cs.ubc.ca>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
There are a few places where libxl__xs_write is passed a variable
value to write to xenstore, but the semantics are that the first char*
is a format string. So use "%s".
This fixes the following errors reported by some newer compilers:
libxl.c: In function "libxl_create_cpupool":
libxl.c:3981: error: format not a string literal and no format arguments
libxl.c:3983: error: format not a string literal and no format arguments
libxl.c: In function "libxl_cpupool_movedomain":
libxl.c:4095: error: format not a string literal and no format arguments
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Gianni Tedesco <gianni.tedesco@citrix.com>
libxl_set_memory_target tries to set the memory target for a given
domain to new_target_memkb. The function includes a check to make sure
that dom0's memory is not reduced below a minimal threshold ie.
LIBXL_MIN_DOM0_MEM. However, this check is performed before the
new_target_memkb variable is properly initialised - when the value is
always zero. This means that the check always fails. Fix this by moving
the test to happen after the proper initialisation of new_target_memkb.
Signed-off-by: Gianni Tedesco <gianni.tedesco@citrix.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
gdbsx hacks CFLAGS to remove optimisations for easier debugging, however
this breaks dep generation due to an early evaluation of CFLAGS and
leaves stale files called "..d" after a make clean. We could possibly
fix this by separating dep generation from the rest of CFLAGS but it
seems a little invasive. The gdbsx hackers can do their own thing during
development and testing anyway...
Signed-off-by: Gianni Tedesco <gianni.tedesco@citrix.com>
When trying to live migrate guest with great memory (e.g. 20 GB),
there's delay caused by destroying source copy of domain. To speed up
this process we can close socket before starting guest destroy, as
source guest is not running anymore.
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
kdd runs in dom0, attaches to a domain and speaks the Windows kd serial
protocol over a TCP connection (which should go to kd or windbg, e.g.
by having another VM with its virtual COM1 set up as a TCP listener).
It doesn't do breakpoints &c yet, and windbg can get quite confused
since the kernel debugger's not actually running, but it's good enough
to extract backtraces from wedged VMs.
Signed-off-by: Tim Deegan <Tim.Deegan@citrix.com>
The callback return convention follows xc_domain_save and therefore
returns 0 on success and >0 on error, correct the check accordingly.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
... so that once in a while Xen knows of a new CPU feature before
Linux starts making use of it.
While (obviously) I wasn't able to test this, it seemed strait forward
enough to enable anyway.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
c/s 13829 lost the (32-bit only) cpu_has_pae argument passed to the
primary kernel's stub (in the 32-bit Xen case only), and Linux
2.6.27/.30 (32-/64-bit) introduced a new argument (for KEXEC_JUMP)
which for now simply gets passed a hardcoded value.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
The memory allocation code sometimes needs to enforce that a guest
that's been told to balloon down isn't going to expand further
(because it's still executing a previous balloon-up operation). That
means being able to set the desired max_pages even before the balloon
driver has brought tot_pages down to the right level.
Signed-off-by: Tim Deegan <Tim.Deegan@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
The current cpupools code interface is a bit inconsistent. This
patch addresses this by making the interaction for each
vcpu in a pool look like this:
alloc_vdata() -- allocates and sets up vcpu data
insert_vcpu() -- the vcpu is ready to run in this pool
remove_vcpu() -- take the vcpu out of the pool
free_vdata() -- delete allocated vcpu data
(Previously, remove_vcpu and free_vdata were combined into a "destroy
vcpu", and insert_vcpu was only called for idle vcpus.)
This also addresses a bug in credit2 which was caused by a
misunderstanding of the cpupools interface.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Acked-by: Juergen Gross <juergen.gross@ts.fujitsu.com>
On restore:
region_mfn is passed to xc_map_foreign_range and
xc_map_foreign_bulk. In both cases the buffer is accessed from the
ioctl handler in the kernel and not from any hypercall. Therefore
normal copy_{to,from}_user handling in the kernel will cope with any
faulting access.
p2m_batch is passed to xc_domain_memory_populate_physmap which takes
care of bouncing the buffer already.
On save:
pfn_type is passed to xc_map_foreign_bulk which does not need locking
as per region_mfn above.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Ian Jackson <ian.jackson.citrix.com>
I have a suspicion these routines should be using libxc rather than
reimplementing all the hypercalls again, but I don't have the
enthusiasm to fix it.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Ian Jackson <ian.jackson.citrix.com>
There is no need to lock/bounce minfo->pfn_type in init_mem_info since
xc_get_pfn_type_batch() will take care of that for us.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: Ian Jackson <ian.jackson.citrix.com>
Keir Fraser
Ian Jackson
Juergen Gross
Ian Campbell
Patrick Colp
Gianni Tedesco
Miroslav Rezanina
Tim Deegan