MMIO exits are more expensive in KVM or Xen than in QEMU because they
involve, at least, privilege transitions. However, MMIO write
operations can be effectively batched if those writes do not have side
effects.
Good examples of this include VGA pixel operations when in a planar
mode. As it turns out, we can get a nice boost in other areas too.
Laurent mentioned a 9.7% performance boost in iperf with the coalesced
MMIO changes for the e1000 when he originally posted this work for KVM.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5961 c046a42c-6fe2-441c-8c8c-71466251a162
Paul's comment on my first approach to fix the h2g usage in
page_find_alloc finally open my eyes about what the code is actually
supposed to do:
With the help of h2g_valid we can no cleanly check if a freshly allocate
page (for host usage) is guest-reachable and, in case it is, mark it
reserved in the guest's address range.
Signed-off-by: Jan Kiszka <jan.kiszka@web.de>
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Edgar E. Iglesias <edgar.iglesias@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5957 c046a42c-6fe2-441c-8c8c-71466251a162
Unfortunately this range is so narrow that I'm not sure if it makes more
sense to always use memory load to pc kind of branch instead.
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5844 c046a42c-6fe2-441c-8c8c-71466251a162
This switches cpu_break/watchpoint_* to TAILQ wrappers, simplifying the
code and also fixing a use after release issue in
cpu_break/watchpoint_remove_all.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5799 c046a42c-6fe2-441c-8c8c-71466251a162
Hypervisors like KVM perform badly while doing mmio on
a loop, because it'll generate an exit on each access.
This is the case with VGA, which results in very bad
performance.
In this patch, we map the linear frame buffer as RAM,
make sure it has dirty region tracking enabled, and then
just let the region to be written.
Cleanups suggestions by:
Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Glauber Costa <glommer@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5793 c046a42c-6fe2-441c-8c8c-71466251a162
ENOBUFS is not defined on Win32. Use ENOMEM instead which is more portable.
This was reported by Hervé Poussineau.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5749 c046a42c-6fe2-441c-8c8c-71466251a162
Add another breakpoint/watchpoint type to BP_GDB: BP_CPU. This type is
intended for hardware-assisted break/watchpoint emulations like the x86
architecture requires.
To keep the highest priority for BP_GDB breakpoints, this type is
always inserted at the head of break/watchpoint lists, thus is found
first when looking up the origin of a debug interruption.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5746 c046a42c-6fe2-441c-8c8c-71466251a162
When one watchpoint is hit, others might have triggered as well. To
support users of the watchpoint API which need to detect such cases,
the BP_WATCHPOINT_HIT flag is introduced and maintained.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5744 c046a42c-6fe2-441c-8c8c-71466251a162
Now that we can properly restore the pc on watchpoint hits, there is no
more need for prematurely terminating TBs if watchpoints are present.
Remove all related bits.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5742 c046a42c-6fe2-441c-8c8c-71466251a162
In order to provide accurate information about the triggering
instruction, this patch adds the required bits to restore the pc if the
access happened inside a TB. With the BP_STOP_BEFORE_ACCESS flag, the
watchpoint user can control if the debug trap should be issued on or
after the accessing instruction.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5741 c046a42c-6fe2-441c-8c8c-71466251a162
This adds length support for watchpoints. To keep things simple, only
aligned watchpoints are accepted.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5740 c046a42c-6fe2-441c-8c8c-71466251a162
This patch prepares the QEMU cpu_watchpoint/breakpoint API to allow the
succeeding enhancements this series comes with.
First of all, it overcomes MAX_BREAKPOINTS/MAX_WATCHPOINTS by switching
to dynamically allocated data structures that are kept in linked lists.
This also allows to return a stable reference to the related objects,
required for later introduced x86 debug register support.
Breakpoints and watchpoints are stored with their full information set
and an additional flag field that makes them easily extensible for use
beyond pure guest debugging.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5738 c046a42c-6fe2-441c-8c8c-71466251a162
This patch refactors the way the CPU state is handled that is associated
with a TB. The basic motivation is to move more arch specific code out
of generic files. Specifically the long #ifdef clutter in tb_find_fast()
has to be overcome in order to avoid duplicating it for the gdb
watchpoint fixes (patch "Restore pc on watchpoint hits").
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5736 c046a42c-6fe2-441c-8c8c-71466251a162
Unfortunately, -linux-user doesn't use osdep as it replaces some of those
functions with specific ones. The code #ifdef code in exec.c needs to
remain in place so instead of introducing a qemu_getpagesize() let's just
use getpagesize() in the non-Windows implementation of qemu_vmalloc.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5703 c046a42c-6fe2-441c-8c8c-71466251a162
This patch adds very basic KVM support. KVM is a kernel module for Linux that
allows userspace programs to make use of hardware virtualization support. It
current supports x86 hardware virtualization using Intel VT-x or AMD-V. It
also supports IA64 VT-i, PPC 440, and S390.
This patch only implements the bare minimum support to get a guest booting. It
has very little impact the rest of QEMU and attempts to integrate nicely with
the rest of QEMU.
Even though this implementation is basic, it is significantly faster than TCG.
Booting and shutting down a Linux guest:
w/TCG: 1:32.36 elapsed 84% CPU
w/KVM: 0:31.14 elapsed 59% CPU
Right now, KVM is disabled by default and must be explicitly enabled with
-enable-kvm. We can enable it by default later when we have had better
testing.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5627 c046a42c-6fe2-441c-8c8c-71466251a162
Move up the warp around test because line
'end = TARGET_PAGE_ALIGN(start+len);'
can interfere with it.
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5563 c046a42c-6fe2-441c-8c8c-71466251a162
This patch adds a dirty tracking bit for live migration. We use 0x08 because
kqemu uses 0x04.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5433 c046a42c-6fe2-441c-8c8c-71466251a162
Don't truncate code_gen_buffer_size calculation to int, as it will give
unpredicted results on 64 bit systems when booting large guests.
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5310 c046a42c-6fe2-441c-8c8c-71466251a162
On some cases, such as under KVM, tb_invalidate_phys_page_range()
may be called for large addresses, when qemu is configured to more than
4GB of RAM.
On these cases, qemu was crashing because it was using an index too
large for l1_map[], that supports only 32-bit addresses when compiling
without CONFIG_USER_ONLY.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@5227 c046a42c-6fe2-441c-8c8c-71466251a162
Save and restore env->interrupt_request and env->halted.
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4817 c046a42c-6fe2-441c-8c8c-71466251a162
The IO index is now stored in its own field, instead of being wedged
into the vaddr field. This eliminates the ROMD and watchpoint host
pointer weirdness. The IO index space is expanded by 1 bit, and
several additional bits are made available in the TLB vaddr field.
git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@4704 c046a42c-6fe2-441c-8c8c-71466251a162
