After allocating 32MB or more contiguous memory, huge pages
would seem to be ideal.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Define a new global dma_context_memory which is a DMAContext corresponding
to the global address_space_memory AddressSpace. This can be used by
sysbus peripherals like sysbus-ohci which need to do DMA.
In particular, use it in the sysbus-ohci device, which fixes a
segfault when attempting to use that device.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
* 'trivial-patches' of git://github.com/stefanha/qemu:
pc: Drop redundant test for ROM memory region
exec: make some functions static
target-ppc: make some functions static
ppc: add missing static
vnc: add missing static
vl.c: add missing static
target-sparc: make do_unaligned_access static
m68k: Return semihosting errno values correctly
cadence_uart: More debug information
Conflicts:
target-m68k/m68k-semi.c
Add GETPC_EXT which is used by MMU helpers to selectively calculate the code
address of accessing guest memory when called from a qemu_ld/st optimized code
or a C function. Currently, it supports only i386 and x86-64 hosts.
Signed-off-by: Yeongkyoon Lee <yeongkyoon.lee@samsung.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Change return type to bool, move to include/qemu/cpu.h and
add documentation.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
[AF: Updated new caller qemu_in_vcpu_thread()]
target_phys_addr_t is unwieldly, violates the C standard (_t suffixes are
reserved) and its purpose doesn't match the name (most target_phys_addr_t
addresses are not target specific). Replace it with a finger-friendly,
standards conformant hwaddr.
Outstanding patchsets can be fixed up with the command
git rebase -i --exec 'find -name "*.[ch]"
| xargs s/target_phys_addr_t/hwaddr/g' origin
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This makes it possible for QEMU to use transparent huge pages (THP)
when transparent_hugepage/enabled=madvise. Otherwise THP is only
used when it's enabled system wide.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
* quintela/migration-next-20121017: (41 commits)
cpus: create qemu_in_vcpu_thread()
savevm: make qemu_file_put_notify() return errors
savevm: un-export qemu_file_set_error()
block-migration: handle errors with the return codes correctly
block-migration: Switch meaning of return value
block-migration: make flush_blks() return errors
buffered_file: buffered_put_buffer() don't need to set last_error
savevm: Only qemu_fflush() can generate errors
savevm: make qemu_fill_buffer() be consistent
savevm: unexport qemu_ftell()
savevm: unfold qemu_fclose_internal()
savevm: make qemu_fflush() return an error code
savevm: Remove qemu_fseek()
virtio-net: use qemu_get_buffer() in a temp buffer
savevm: unexport qemu_fflush
migration: make migrate_fd_wait_for_unfreeze() return errors
buffered_file: make buffered_flush return the error code
buffered_file: callers of buffered_flush() already check for errors
buffered_file: We can access directly to bandwidth_limit
buffered_file: unfold migrate_fd_close
...
* qemu-kvm/memory/dma: (23 commits)
pci: honor PCI_COMMAND_MASTER
pci: give each device its own address space
memory: add address_space_destroy()
dma: make dma access its own address space
memory: per-AddressSpace dispatch
s390: avoid reaching into memory core internals
memory: use AddressSpace for MemoryListener filtering
memory: move tcg flush into a tcg memory listener
memory: move address_space_memory and address_space_io out of memory core
memory: manage coalesced mmio via a MemoryListener
xen: drop no-op MemoryListener callbacks
kvm: drop no-op MemoryListener callbacks
xen_pt: drop no-op MemoryListener callbacks
vfio: drop no-op MemoryListener callbacks
memory: drop no-op MemoryListener callbacks
memory: provide defaults for MemoryListener operations
memory: maintain a list of address spaces
memory: export AddressSpace
memory: prepare AddressSpace for exporting
xen_pt: use separate MemoryListeners for memory and I/O
...
Currently we use a global radix tree to dispatch memory access. This only
works with a single address space; to support multiple address spaces we
make the radix tree a member of AddressSpace (via an intermediate structure
AddressSpaceDispatch to avoid exposing too many internals).
A side effect is that address_space_io also gains a dispatch table. When
we remove all the pre-memory-API I/O registrations, we can use that for
dispatching I/O and get rid of the original I/O dispatch.
Signed-off-by: Avi Kivity <avi@redhat.com>
Using the AddressSpace type reduces confusion, as you can't accidentally
supply the MemoryRegion you're interested in.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
We plan to make the core listener listen to all address spaces; this
will cause many more flushes than necessary. Prepare for that by
moving the flush into a tcg-specific listener.
Later we can avoid registering the listener if tcg is disabled.
Signed-off-by: Avi Kivity <avi@redhat.com>
With this change, memory.c no longer knows anything about special address
spaces, so it is prepared for AddressSpace based DMA.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Instead of calling a global function on coalesced mmio changes, which
routes the call to kvm if enabled, add coalesced mmio hooks to
MemoryListener and make kvm use that instead.
The motivation is support for multiple address spaces (which means we
we need to filter the call on the right address space) but the result
is cleaner as well.
Signed-off-by: Avi Kivity <avi@redhat.com>
It is used nowhere else, and the corresponding MAX_CODE_GEN_BUFFER_SIZE
also lives there.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
We had a hack for arm and sparc, allocating code_gen_prologue to a
special section. Which, honestly does no good under certain cases.
We've already got limits on code_gen_buffer_size to ensure that all
TBs can use direct branches between themselves; reuse this limit to
ensure the prologue is also reachable.
As a bonus, we get to avoid marking a page of the main executable's
data segment as executable.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The hard-coded addresses inside alloc_code_gen_buffer only make sense
if we're building an executable that will actually run at the address
we've put into the linker scripts.
When we're building with -fpie, the executable will run at some
random location chosen by the kernel. We get better placement for
the code_gen_buffer if we allow the kernel to place the memory,
as it will tend to to place it near the executable, based on the
PROT_EXEC bit.
Since code_gen_prologue is always inside the executable, this effect
is easily seen at the end of most TB, with the exit_tb opcode, and
with any calls to helper functions.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
For ARM we cap the buffer size to 16MB. Do not allocate 32MB in that case.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
It now consists of:
A macro definition of MAX_CODE_GEN_BUFFER_SIZE with host-specific values,
A function size_code_gen_buffer that applies most of the reasoning for
choosing a buffer size,
Three variations of a function alloc_code_gen_buffer that contain all
of the logic for allocating executable memory via a given allocation
mechanism.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
exec-obsolete.h used to hold pre-memory-API functions that were used from
device code prior to the transition to the memory API. Now that the
transition is complete, the name no longer describes the file. The
functions still need to be merged better into the memory core, but there's
no danger of anyone using them.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Move the DUMP_FPU and DUMP_CCOP flags for cpu_dump_state() from being
x86-specific flags to being generic ones. This allows us to drop some
TARGET_I386 ifdefs in various places, and means that we can (potentially)
be more consistent across architectures about which monitor commands or
debug abort printouts include FPU register contents and info about
QEMU's condition-code optimisations.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This patch add some calls to xen_modified_memory to notify Xen about dirtybits
during migration.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Avi Kivity <avi@redhat.com>
This new helper/hook is used in the next patch to add an extra call in a single
place.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Reviewed-by: Avi Kivity <avi@redhat.com>
Current code doesn't actually work in 32-bit mode at all. Since
no one really noticed, drop the complication of v7 and v8 cpus.
Eliminate the --sparc_cpu configure option and standardize macro
testing on TCG_TARGET_REG_BITS / HOST_LONG_BITS
Signed-off-by: Richard Henderson <rth@twiddle.net>
The address we pick in sparc64.ld is also 0x60000000, so doing a fixed map
on top of that is guaranteed to blow up. Choosing 0x40000000 is exactly
right for the max of code_gen_buffer_size set below.
No need to ever use MAP_FIXED. While getting our desired address helps
optimize the generated code, we won't fail if we don't get it.
Signed-off-by: Richard Henderson <rth@twiddle.net>
cpu_physical_memory_write_rom(), despite the name, can also be used to
write images into RAM - and will often be used that way if the machine
uses load_image_targphys() into RAM addresses.
However, cpu_physical_memory_write_rom(), unlike cpu_physical_memory_rw()
doesn't invalidate any cached TBs which might be affected by the region
written.
This was breaking reset (under full emu) on the pseries machine - we loaded
our firmware image into RAM, and while executing it rewrite the code at
the entry point (correctly causing a TB invalidate/refresh). When we
reset the firmware image was reloaded, but the TB from the rewrite was
still active and caused us to get an illegal instruction trap.
This patch fixes the bug by duplicating the tb invalidate code from
cpu_physical_memory_rw() in cpu_physical_memory_write_rom().
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
It allows to disable memory merge support (KSM on Linux), which is
enabled by default otherwise.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Add a new '[,dump-guest-core=on|off]' option to the '-machine' option. When
'dump-guest-core=off' is specified, guest memory is omitted from the core dump.
The default behavior continues to be to include guest memory when a core dump is
triggered. In my testing, this brought the core dump size down from 384MB to 6MB
on a 2GB guest.
Is anything additional required to preserve this setting for migration or
savevm? I don't believe so.
Changelog:
v3:
Eliminate globals as per Anthony's suggestion
set no dump from qemu_ram_remap() as well
v2:
move the option from -m to -machine, rename option dump -> dump-guest-core
Signed-off-by: Jason Baron <jbaron@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
For each newly created RAM block, dirty bitmap is reallocated with g_realloc, which doesn't
make any promises on initial content of new extra data in returned buffer. In theory,
we initialize this new data with cpu_physical_memory_set_dirty_range() call. The
problem is, cpu_physical_memory_set_dirty_range() has a side effect of incrementing
ram_list.dirty_pages variable, but only for pages which are not already dirty. And
page "cleanliness" is determined using the same not yet uninitialized dirty bitmap
we've just reallocated. This results in inconsistency between real dirty page number
and value in ram_list.dirty_pages variable, which in turn could (and will) result
in errors during VM migration.
Zero initialize new dirty bitmap bytes to fix this problem.
Signed-off-by: Igor Mitsyanko <i.mitsyanko@samsung.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Remove an out of date comment: this comment used to be attached to
cpu_register_physical_memory_log(), before commit 0f0cb164 accidentally
inserted a couple of other functions between the comment and its function.
It is in any case obsolete since (a) the function arguments it refers
to have been replaced with a single MemoryRegionSection* argument and
(b) the inability to handle regions whose offset_within_address_space
and offset_within_region aren't equally aligned was fixed as part of
the rewrite of this code.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Registering a multi-page memory region that is non-page-aligned results
in a subpage from the start to the page boundary, some number of full
pages, and possibly another subpage from the last page boundary to the
end. The full pages will have a value for offset_within_region that is
not a multiple of TARGET_PAGE_SIZE. Accesses through softmmu are unable
to handle this and will segfault.
Handling full pages through subpages is not optimal, but only
non-page-aligned mappings take the penalty.
Signed-off-by: Tyler Hall <tylerwhall@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
subpage_register() expects "end" to be the last byte in the mapping.
Registering a non-page-aligned memory region that extends up to or
beyond a page boundary causes subpage_register() to silently fail
through the (end >= PAGE_SIZE) check.
This bug does not cause noticeable problems for mappings that do not
extend to a page boundary, though they do register an extra byte.
Signed-off-by: Tyler Hall <tylerwhall@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
* qemu-kvm/uq/master:
virtio: move common irqfd handling out of virtio-pci
virtio: move common ioeventfd handling out of virtio-pci
event_notifier: add event_notifier_set_handler
memory: pass EventNotifier, not eventfd
ivshmem: wrap ivshmem_del_eventfd loops with transaction
ivshmem: use EventNotifier and memory API
event_notifier: add event_notifier_init_fd
event_notifier: remove event_notifier_test
event_notifier: add event_notifier_set
apic: Defer interrupt updates to VCPU thread
apic: Reevaluate pending interrupts on LVT_LINT0 changes
apic: Resolve potential endless loop around apic_update_irq
kvm: expose tsc deadline timer feature to guest
kvm_pv_eoi: add flag support
kvm: Don't abort on kvm_irqchip_add_msi_route()
Under Win32, EventNotifiers will not have event_notifier_get_fd, so we
cannot call it in common code such as hw/virtio-pci.c. Pass a pointer to
the notifier, and only retrieve the file descriptor in kvm-specific code.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
By default qemu will use MAP_PRIVATE for guest pages. This will write
protect pages and thus break on s390 systems that dont support this feature.
Therefore qemu has a hack to always use MAP_SHARED for s390. But MAP_SHARED
has other problems (no dirty pages tracking, a lot more swap overhead etc.)
Newer systems allow the distinction via KVM_CAP_S390_COW. With this feature
qemu can use the standard qemu alloc if available, otherwise it will use
the old s390 hack.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Jens Freimann <jfrei@linux.vnet.ibm.com>
Acked-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Refactor the code that is only needed for tcg to an static function.
Call that only when tcg is enabled. We can't refactor to a dummy
function in the kvm case, as qemu can be compiled at the same time
with tcg and kvm.
Signed-off-by: Juan Quintela <quintela@redhat.com>
This makes it easier to remove it from BusInfo.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[AF: Drop now unnecessary NULL initialization in scsibus_get_dev_path()]
Signed-off-by: Andreas Färber <afaerber@suse.de>
tb_invalidate_phys_addr has to be called with the exact physical address of
the breakpoint we add/remove, not just the page's base address.
Otherwise we easily fail to flush the right TB.
This breakage was introduced by the commit f3705d5329 "memory: make
phys_page_find() return an unadjusted".
This appeared to work for some guest architectures because their
cpu_get_phys_page_debug implementation returns full translated physical
address, not just the base of the TARGET_PAGE_SIZE-sized page.
Reported-by: TeLeMan <geleman@gmail.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
They could suggest that all TBs of the page containing the range would
be invalidated.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
This API will be used in the following patch.
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
If we execute linux-user code that does the following:
* A = mmap()
* execute code in A
* munmap(A)
* B = mmap(), but mmap returns the same address as A
* execute code in B
we end up executing a stale cached tb that contains translated code
from A, while we want new code from B.
This patch adds a TB flush for mmap'ed regions, before we return them,
avoiding the whole issue. It also adds a flush for munmap, so that we
don't execute stale TBs instead of getting a segfault.
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Acked-by: Riku Voipio <riku.voipio@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Fold is_ram_rom and is_ram_rom_romd() into callers.
Change is_romd() and section_addr() to take MemoryRegion
instead of MemoryRegionSection for consistency and
use memory_region_ prefix.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Make s_cputlb_empty_entry 'const'.
Rename tlb_flush_jmp_cache() to tb_flush_jmp_cache().
Refactor code to add cpu_tlb_reset_dirty_all(),
memory_region_section_get_iotlb() and
memory_region_is_unassigned().
Remove unused cpu_tlb_update_dirty().
Fix coding style in areas to be moved.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Replace all type casts to 'long' or 'unsigned long' by 'intptr_t' or 'uintptr_t'.
For type casts which are only used to extract the lower bits of an address
or to modify those bits, signedness does not matter. There I always use 'uintptr_t'.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Allow TB invalidation by its physical address, extract implementation
from the breakpoint_invalidate function.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Use uintptr_t instead of void * or unsigned long in
several op related functions, env->mem_io_pc and
GETPC() macro.
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
QEMU host addresses must use uintptr_t to be portable for hosts with
an unusual size of long (w64).
tb_jmp_offset is an uint16_t value, therefore the local variable offset
in function tb_set_jmp_target was changed from unsigned long to uint16_t.
The type cast to long in function tb_add_jump now also uses uintptr_t.
For the bit operation used here, the signedness of the type cast does
not matter.
Some remaining unsigned long values are either only used for ARM assembler
code or will be fixed in a later patch for PPC.
v2:
Fix signature of tb_find_pc in exec.c, too (hint from Blue Swirl, thanks).
There remain lots of other long / unsigned long in exec.c which must be
replaced by uintptr_t. This will be done in a separate patch. Here
only one of these type casts is fixed.
v3:
Also fix signature of page_unprotect.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
This allows us to generate unwind info for the dynamicly generated
code in the code_gen_buffer. Only i386 is converted at this point.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
In cpu_physical_memory_rw, a change has been introduced and qemu_get_ram_ptr is
no longuer called with the ram addr we want to access, but only with the
section address. This patch fixes this. (All other call to qemu_get_ram_ptr are
already called with the right address.)
This patch fixes Xen guest.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
The code to get the ram_addr from a (tlb entry, vaddr) pair
checks that the resulting memory is not MMIO, but neglects to
check whether the region is hidden by a watchpoint page.
Add the missing check.
Signed-off-by: Avi Kivity <avi@redhat.com>
A couple of code paths check the lower bits of CPUTLBEntry::addr_write
against io_mem_ram as a way of looking for a dirty RAM page. This works
by accident since the value is zero, which matches all clear bits for
TLB_INVALID, TLB_MMIO, and TLB_NOTDIRTY (indicating dirty RAM).
Make it work by design by checking for the proper bits.
Signed-off-by: Avi Kivity <avi@redhat.com>
Optionally, make memory access helpers take a parameter for CPUState
instead of relying on global env.
On most targets, perform simple moves to reorder registers. On i386,
switch from regparm(3) calling convention to standard stack-based
version.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Scripted conversion:
for file in *.[hc] hw/*.[hc] hw/kvm/*.[hc] linux-user/*.[hc] linux-user/m68k/*.[hc] bsd-user/*.[hc] darwin-user/*.[hc] tcg/*/*.[hc] target-*/cpu.h; do
sed -i "s/CPUState/CPUArchState/g" $file
done
All occurrences of CPUArchState are expected to be replaced by QOM CPUState,
once all targets are QOM'ified and common fields have been extracted.
Signed-off-by: Andreas Färber <afaerber@suse.de>
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
The return value of cpu_register_io_memory() is no longer used anywhere, so
we can remove it and all associated data and code.
Signed-off-by: Avi Kivity <avi@redhat.com>
Instead of indirecting via io_mem_region, dispatch directly
through the MemoryRegion obtained from the iotlb or phys_page_find().
Signed-off-by: Avi Kivity <avi@redhat.com>
get_page_addr_code() reads a code tlb entry, but interprets it as an
iotlb entry. This works by accident since the low bits of a RAM code
tlb entry are clear, and match a RAM iotlb entry. This accident is
about to unhappen, so fix the code to use an iotlb entry (using the
code entry with TLB_MMIO may fail if the page is a watchpoint).
Signed-off-by: Avi Kivity <avi@redhat.com>
We'd like to store the section index in the iotlb, so we can't
adjust it before returning. Return an unadjusted section and
instead introduce section_addr(), which does the adjustment later.
Signed-off-by: Avi Kivity <avi@redhat.com>
Commit e58ac72b6a0 ("ioport: change portio_list not to use
memory_region_set_offset()") started using aliases of I/O memory
regions. Since the IORange used for the I/O was contained in the
target region, the alias information (specifically, the offset
into the region) was lost. This broke -vga std.
Fix by allocating an independent object to hold the IORange and
also the new offset.
Note that I/O memory regions were conceptually broken wrt aliases
in a different way: an alias can cause the same region to appear
twice in an address space, but we had just one IORange to service it.
This patch fixes that problem as well, since we can now have multiple
IORange/MemoryRegion associations.
Signed-off-by: Avi Kivity <avi@redhat.com>
When storing large contiguous ranges in phys_map, all values tend to
be the same pointers to a single MemoryRegionSection. Collapse them
by marking nodes with level > 0 as leaves. This reduces tree memory
usage dramatically.
Signed-off-by: Avi Kivity <avi@redhat.com>
Instead of considering subpage on a per-page basis, split each section
into a subpage head, multipage body, and subpage tail, and register
each separately. This simplifies the registration functions.
Signed-off-by: Avi Kivity <avi@redhat.com>
We no longer describe memory in terms of individual pages; use sections
throughout instead.
PhysPageDesc no longer used - remove.
Signed-off-by: Avi Kivity <avi@redhat.com>
If the first subpage installed in a page is RAM, then we install it as
a full page, instead of a subpage. Fix by not special casing RAM.
The issue dates to commit db7b5426a4, which introduced subpages.
Signed-off-by: Avi Kivity <avi@redhat.com>
Use an expanding vector to store nodes. Allocation is baroque to g_renew()
potentially invalidating pointers; this will be addressed later.
Signed-off-by: Avi Kivity <avi@redhat.com>
Instead of storing PhysPageDesc, store pointers to MemoryRegionSections.
The various offsets (phys_offset & ~TARGET_PAGE_MASK,
PHYS_OFFSET & TARGET_PAGE_MASK, region_offset) can all be synthesized
from the information in a MemoryRegionSection. Adjust phys_page_find()
to synthesize a PhysPageDesc.
The upshot is that phys_map now contains uniform values, so it's easier
to generate and compress.
The end result is somewhat clumsy but this will be improved as we we
propagate MemoryRegionSections throughout the code instead of transforming
them to PhysPageDesc.
The MemoryRegionSection pointers are stored as uint16_t offsets in an
array. This saves space (when we also compress node pointers) and is
more cache friendly.
Signed-off-by: Avi Kivity <avi@redhat.com>
L1 and the lower levels in l1_phys_map are equivalent, except that L1 has
a different size, and is always allocated. Simplify the code by removing
L1. This leaves us with a tree composed solely of L2 tables, but that
problem can be renamed away later.
Signed-off-by: Avi Kivity <avi@redhat.com>
Instead of incrementally building the memory map, rebuild it every time.
This allows later simplification, since the code need not consider overlaying
a previous mapping. It is also RCU friendly.
With large memory guests this can get expensive, since the operation is
O(mem size), but this will be optimized later.
As a side effect subpage and L2 leaks are fixed here.
Signed-off-by: Avi Kivity <avi@redhat.com>
Current memory listeners are incremental; that is, they are expected to
maintain their own state, and receive callbacks for changes to that state.
This patch adds support for stateless listeners; these work by receiving
a ->begin() callback (which tells them that new state is coming), a
sequence of ->region_add() and ->region_nop() callbacks, and then a
->commit() callback which signifies the end of the new state. They should
ignore ->region_del() callbacks.
Signed-off-by: Avi Kivity <avi@redhat.com>
This transforms memory.c into a library which can then be unit tested
easily, by feeding it inputs and listening to its outputs.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
It can be derived from the MemoryRegion itself (which is why it is not
used there).
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
In case of BP_STOP_BEFORE_ACCESS watchpoint check_watchpoint intends to
signal EXCP_DEBUG exception on exit from cpu loop, but later overwrites
exception code by the cpu_resume_from_signal call.
Use cpu_loop_exit with BP_STOP_BEFORE_ACCESS watchpoints.
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Clarify the comment about tlb_flush()'s flush_global parameter,
so it is clearer what it does and why it is OK that the implementation
currently ignores it.
Reviewed-by: Andreas F=C3=A4rber <afaerber@suse.de>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The virtio config area in PIO space is a bit special. The initial
header is little endian but the rest (device specific) is guest
native endian.
The PIO accessors for PCI on machines that don't have native IO ports
assume that all PIO is little endian, which works fine for everything
except the above.
A complicated way to fix it would be to split the BAR into two memory
regions with different endianess settings, but this isn't practical
to do, besides, the PIO code doesn't honor region endianness anyway
(I have a patch for that too but it isn't necessary at this stage).
So I decided to go for the quick fix instead which consists of
reverting the swap in virtio-pci in selected places, hoping that when
we eventually do a "v2" of the virtio protocols, we sort that out once
and for all using a fixed endian setting for everything.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
[agraf: keep virtio in libhw and determine endianness through a
helper function in exec.c]
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
ARM still doesn't support 16GB buffers in 32-bit modes, replace the
16GB by 16MB in the comment.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
We no longer use any of the lower bits of a ram_addr, so we might as well
use them for the io table index. This increases the number of potential
I/O handlers by a factor of 8.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Unlike ->readonly, ->readable is not inherited from aliase, so we can simply
query the memory region.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Now that all mmio goes through MemoryRegions, we can convert
io_mem_opaque to be a MemoryRegion pointer, and remove the thunks
that convert from old-style CPU{Read,Write}MemoryFunc to MemoryRegionOps.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Convert the fixed-address IO_MEM_RAM, IO_MEM_ROM, IO_MEM_UNASSIGNED,
and IO_MEM_NOTDIRTY io handlers to MemoryRegions. These aren't real
regions, since they are never added to the memory hierarchy, but they
allow reuse of the dispatch functionality.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Its use of IO_MEM_ROM and friends will later cause #include loops; and it
is too large to merit inlining.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
The code sometimes uses range comparisons on io indexes (e.g.
index =< IO_MEM_ROM). Avoid these as they make moving to objects harder.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
cpu_register_physical_memory_log() does not update region_offset
if a page was previously registered for the same address. This
could cause mmio accesses going to the wrong place, by using the
old region_offset.
Signed-off-by: Avi Kivity <avi@redhat.com>
Acked-by: Andreas Färber <afaerber@suse.de>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Currently mmio access goes directly to the io_mem_{read,write} arrays.
In preparation for eliminating them, add indirection via a function.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Instead of returning a PhysPageDesc pointer, return a temporary.
This lets us move away from actually storing PhysPageDesc's, and
instead sythesising them when needed.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Instead of doing device endianness compensation in cpu_register_io_memory(),
do it in the memory core.
Signed-off-by: Avi Kivity <avi@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
The getter is no longer used, so it is completely removed.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
As a step in moving live migration from RAMBlocks to MemoryRegions,
store the MemoryRegion in a RAMBlock.
Reviewed-by: Anthony Liguori <aliguori@us.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Currently creating a memory region automatically registers it for
live migration. This differs from other state (which is enumerated
in a VMStateDescription structure) and ties the live migration code
into the memory core.
Decouple the two by introducing a separate API, vmstate_register_ram(),
for registering a RAM block for migration. Currently the same
implementation is reused, but later it can be moved into a separate list,
and registrations can be moved to VMStateDescription blocks.
Signed-off-by: Avi Kivity <avi@redhat.com>
Add an API that allows a client to observe changes in the global
memory map:
- region added (possibly with logging enabled)
- region removed (possibly with logging enabled)
- logging started on a region
- logging stopped on a region
- global logging started
- global logging removed
This API will eventually replace cpu_register_physical_memory_client().
Signed-off-by: Avi Kivity <avi@redhat.com>
Currently xen_ram_alloc() relies on ram_addr, which is going away.
Give it something else to use as a cookie.
Signed-off-by: Avi Kivity <avi@redhat.com>
This fixes a common bug with initial region_offset value.
Usually, the pages are re-assigned afterwards, so the bug
has a very small effect on regular QEMU use flows.
Signed-off-by: Alex Rozenman <Alex_Rozenman@mentor.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Commit 95c318f5e1 (Fix segfault in mmio
subpage handling code.) prevented a segfault by making all subpage
registrations over an existing memory page perform an unassigned access.
Symptoms were writes not taking effect and reads returning zero.
Very small page sizes are not currently supported either,
so subpage memory areas cannot fully be avoided.
Therefore change the previous fix to use a new IO_MEM_SUBPAGE_RAM
instead of IO_MEM_UNASSIGNED. Suggested by Avi.
Reviewed-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Cc: Avi Kivity <avi@redhat.com>
Cc: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
On ARM, don't map the code buffer at a fixed location, and fix up the
call/goto tcg routines to let it do long jumps.
Mapping the code buffer at a fixed address could sometimes result in it being
mapped over the top of the heap with pretty random results.
Signed-off-by: Dr. David Alan Gilbert <david.gilbert@linaro.org>
Signed-off-by: Andrzej Zaborowski <andrew.zaborowski@intel.com>
W32 does not support line buffering, but it supports unbuffered output.
Unbuffered output is better for writing to qemu.log than fully buffered
output because it also shows the latest log messages when an application
crash occurs.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Make cpu_single_env thread-local. This fixes a regression
in handling of multi-threaded programs in linux-user mode
(bug 823902).
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[Peter Maydell: rename tls_cpu_single_env to cpu_single_env]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Spotted via code review, we initialize offset to 0 to avoid a
compiler warning, but in the unlikely case that offset is
never set to something else, we should abort instead of return
a value that will almost certainly cause problems.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
As phys_ram_size had been removed since QEMU 0.12. Remove the useless
comment.
Signed-off-by: Chen Wen-Ren <chenwj@iis.sinica.edu.tw>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
GETPC() can be used even from outside of helper code. Move the macro to
a more accessible location. Avoid a compile warning from redefining it in exec.c.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
It was introduced with commit 54936004fd
as host_page_bits but never used.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Spotted while reviewing the migration thread patches.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
get_system_io() returns the root I/O memory region.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Use mmap to allocate executable memory on NetBSD as well.
Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Do not allocate TCG-only resources like the translation buffer when
running over KVM or XEN. Saves a "few" bytes in the qemu address space
and is also conceptually cleaner.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
When trying to map an alias of a ram region, where the alias starts at
address A and we map it into address B, and A > B, we had an arithmetic
underflow. Because we use unsigned arithmetic, the underflow converted
into a large number which failed addrrange_intersects() tests.
The concrete example which triggered this was cirrus vga mapping
the framebuffer at offsets 0xc0000-0xc7fff (relative to the start of
the framebuffer) into offsets 0xa0000 (relative to system addres space
start).
With our favorite analogy of a windowing system, this is equivalent to
dragging a subwindow off the left edge of the screen, and failing to clip
it into its parent window which is on screen.
Fix by switching to signed arithmetic.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
In Xen case, memory can be bigger than the host memory. that mean a
32bits host (and QEMU) should be able to handle a RAM address of 64bits.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
As the variable pd and addr1 inside the function cpu_physical_memory_rw
are mean to handle a RAM address, they should be of the ram_addr_t type
instead of unsigned long.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
cea5f9a28f exposed bugs in unassigned memory
access handling. Fix them by always passing CPUState to the handlers.
Reported-by: Hervé Poussineau <hpoussin@reactos.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
qemu_ram_ptr_length should take ram_addr_t as argument rather than
target_phys_addr_t because is doing comparisons with RAMBlock addresses.
cpu_physical_memory_map should create a ram_addr_t address to pass to
qemu_ram_ptr_length from PhysPageDesc phys_offset.
Remove code after abort() in qemu_ram_ptr_length.
Changes in v2:
- handle 0 size in qemu_ram_ptr_length;
- rename addr1 to raddr;
- initialize raddr to ULONG_MAX.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Xen won't be enabled if there is no backend support available for the
host. And that also means the map cache will work. So drop the separate
config switch and move the required stubs over to xen-stub.c.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
The map cache is a Xen thing, so its API should make this clear.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
When calculating the point at which we should not try to put another
TB into the code gen buffer, we have to allow not just for OPC_MAX_SIZE
but OPC_BUF_SIZE. This is because the target translate.c will only
stop when an instruction has put it past the OPC_MAX_SIZE limit, so
we have to include the MAX_OP_PER_INSTR margin which that final insn
might have used.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Device code some times needs to access physical memory and does that
through the ld./st._phys functions. However, these are the exact same
functions that the CPU uses to access memory, which means they will
be endianness swapped depending on the target CPU.
However, devices don't know about the CPU's endianness, but instead
access memory directly using their own interface to the memory bus,
so they need some way to read data with their native endianness.
This patch adds _le and _be functions to ld./st._phys.
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Once there, use a better variable name.
Signed-off-by: Juan Quintela <quintela@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Use qemu_invalidate_entry in cpu_physical_memory_unmap.
Do not lock mapcache entries in qemu_get_ram_ptr if the address falls in
the ramblock with offset == 0. We don't need to do that because the
callers of qemu_get_ram_ptr either try to map an entire block, other
from the main ramblock, or until the end of a page to implement a single
read or write in the main ramblock.
If we don't lock mapcache entries in qemu_get_ram_ptr we don't need to
call qemu_invalidate_entry in qemu_put_ram_ptr anymore because we can
leave with few long lived block mappings requested by devices.
Also move the call to qemu_ram_addr_from_mapcache at the beginning of
qemu_ram_addr_from_host.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Introduce qemu_ram_ptr_length that takes an address and a size as
parameters rather than just an address.
Refactor cpu_physical_memory_map so that we call qemu_ram_ptr_length only
once rather than calling qemu_get_ram_ptr one time per page.
This is not only more efficient but also tries to simplify the logic of
the function.
Currently we are relying on the fact that all the pages are mapped
contiguously in qemu's address space: we have a check to make sure that
the virtual address returned by qemu_get_ram_ptr from the second call on
is consecutive. Now we are making this more explicit replacing all the
calls to qemu_get_ram_ptr with a single call to qemu_ram_ptr_length
passing a size argument.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
CC: agraf@suse.de
CC: anthony@codemonkey.ws
Signed-off-by: Alexander Graf <agraf@suse.de>
Replace xen_map_block with qemu_map_cache with the appropriate locking
and size parameters.
Replace xen_unmap_block with qemu_invalidate_entry.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
There is no need for qemu_map_cache_unlock, just use
qemu_invalidate_entry instead.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
When a phys memory client registers and we play catchup by walking
the page tables, we can make a huge improvement in the number of
times the set_memory callback is called by batching contiguous
pages together. With a 4G guest, this reduces the number of callbacks
at registration from 1048866 to 296.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
* rth/axp-next: (26 commits)
target-alpha: Implement TLB flush primitives.
target-alpha: Use a fixed frequency for the RPCC in system mode.
target-alpha: Trap for unassigned and unaligned addresses.
target-alpha: Remap PIO space for 43-bit KSEG for EV6.
target-alpha: Implement cpu_alpha_handle_mmu_fault for system mode.
target-alpha: Implement more CALL_PAL values inline.
target-alpha: Disable interrupts properly.
target-alpha: All ISA checks to use TB->FLAGS.
target-alpha: Swap shadow registers moving to/from PALmode.
target-alpha: Implement do_interrupt for system mode.
target-alpha: Add IPRs to be used by the emulation PALcode.
target-alpha: Use kernel mmu_idx for pal_mode.
target-alpha: Add various symbolic constants.
target-alpha: Use do_restore_state for arithmetic exceptions.
target-alpha: Tidy up arithmetic exceptions.
target-alpha: Tidy exception constants.
target-alpha: Enable the alpha-softmmu target.
target-alpha: Rationalize internal processor registers.
target-alpha: Merge HW_REI and HW_RET implementations.
target-alpha: Cleanup MMU modes.
...
This patch removes all references to signal.h when qemu-common.h is included
as they become redundant.
Signed-off-by: Alexandre Raymond <cerbere@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Required for regions mapped via qemu_ram_alloc_from_ptr(). VFIO
and ivshmem will make use of this to remove mappings when devices
are hot unplugged.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
While trying out the > 64GB guest RAM patch, I hit some virtual address
limitations of my host system, which resulted in mmap failing. Unfortunately,
qemu didn't tell me about this failure, but just used the NULL pointer
happily, resulting in either segmentation faults or other fun errors.
To spare other users from tracing this down, let's print a nice message
instead so the user can figure out what's wrong from there.
Signed-off-by: Alexander Graf <agraf@suse.de>
the current s390x qemu memory layout is
0x1000000: guest start
0x80000000: qemu binary
which limits the amount of available memory to <2GB.
This patch moves the guest pages to 32GB to not collide with the binary
and to leave some space for the program break of qemu.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
This function allows to unlock a ram_ptr give by qemu_get_ram_ptr. After
a call to qemu_put_ram_ptr, the pointer may be unmap from QEMU when
used with Xen.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Acked-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Alexander Graf <agraf@suse.de>
On IA32 host or IA32 PAE host, at present, generally, we can't create
an HVM guest with more than 2G memory, because generally it's almost
impossible for Qemu to find a large enough and consecutive virtual
address space to map an HVM guest's whole physical address space.
The attached patch fixes this issue using dynamic mapping based on
little blocks of memory.
Each call to qemu_get_ram_ptr makes a call to qemu_map_cache with the
lock option, so mapcache will not unmap these ram_ptr.
Blocks that do not belong to the RAM, but usually to a device ROM or to
a framebuffer, are handled in a separate function. So the whole RAMBlock
can be map.
Signed-off-by: Jun Nakajima <jun.nakajima@intel.com>
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
When we're trying to get a newly registered phys memory client updated
with the current page mappings, we end up passing the region offset
(a ram_addr_t) as the start address rather than the actual guest
physical memory address (target_phys_addr_t). If your guest has less
than 3.5G of memory, these are coincidentally the same thing. If
there's more, the region offset for the memory above 4G starts over
at 0, so the set_memory client will overwrite it's lower memory entries.
Instead, keep track of the guest phsyical address as we're walking the
tables and pass that to the set_memory client.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When we register a physical memory client, we try to walk the page
tables, calling the set_memory hook for every entry. Effectively
playing catchup for the client for everything already registered.
With this type, we only walk the 2nd entry of the l1 table,
typically missing all of the registered memory.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This allows to override the interrupt handling of QEMU in system mode.
KVM will make use of it to set a specialized handler.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Both have only two lines in common, and we will convert the system
service into a callback which is of no use for user mode operation.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
CC: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
The previous patch removed the need for parameter puc.
Is is now unused, so remove it.
Cc: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Using cpu_physical_memory_read, cpu_physical_memory_write and ldub_phys
improves readability and allows removing some type casts.
lduw_phys and ldl_phys were not used because both require aligned
addresses. Therefore it is not possible to simply replace existing
calls by one of these functions.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
All other type casts in calls of cpu_physical_memory_write are
used by hardware emulations and will be fixed by separate patches.
Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Based on patch by Glauber Costa:
To allow management applications like libvirt to apply CPU affinities to
the VCPU threads, expose their ID via info cpus. This patch provides the
pre-existing and used interface from qemu-kvm.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
This both detects invalid invocations of qemu_ram_free and
qemu_ram_remap when mem_path is non-NULL and fixes a build error on
s390 ("'area' may be used uninitialized in this function").
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
CC: Alexander Graf <agraf@suse.de>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
qemu_ram_remap() unmaps the specified RAM pages, then re-maps these
pages again. This is used by KVM HWPoison support to clear HWPoisoned
page tables across guest rebooting, so that a new page may be
allocated later to recover the memory error.
[ Jan: style fixlets, WIN32 fix ]
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
We have qemu_cpu_self and qemu_thread_self. The latter is retrieving the
current thread, the former is checking for equality (using CPUState). We
also have qemu_thread_equal which is only used like qemu_cpu_self.
This refactors the interfaces, creating qemu_cpu_is_self and
qemu_thread_is_self as well ass qemu_thread_get_self.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
When the commit f471a17e9d converted the
ram_blocks structure to QLIST, it also removed the conditional check before
switching the current block at the beginning of the list.
In the common use case where ram_blocks has a few blocks with only one
frequently accessed (the main RAM), this has a performance impact as it
performs the useless list operations on each call (which are on a really
hot path).
On my machine emulation (ARM on amd64), this patch reduces the
percentage of CPU time spent in qemu_get_ram_ptr from 6.3% to 2.1% in the
profiling of a full boot.
Signed-off-by: Vincent Palatin <vpalatin@chromium.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
In order to use log_start/log_stop with Xen as well in the vga code,
this two operations have been put in CPUPhysMemoryClient.
The two new functions cpu_physical_log_start,cpu_physical_log_stop are
used in hw/vga.c and replace the kvm_log_start/stop. With this, vga does
no longer depends on kvm header.
[ Jan: rebasing and style fixlets ]
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
This function is only used within exec.c, so no need to make it public.
Signed-off-by: Tristan Gingold <gingold@adacore.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
With current OpenBSD, code_gen_buffer was mapped 8GB away from
text segment. Then any helpers were beyond the 2GB range of call
instruction genereated by TCG and so the calls would go nowhere,
leading to a segfault.
Fix by specifying an address for the code_gen_buffer,
hopefully free and nearby the helpers.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
As stated before, devices can be little, big or native endian. The
target endianness is not of their concern, so we need to push things
down a level.
This patch adds a parameter to cpu_register_io_memory that allows a
device to choose its endianness. For now, all devices simply choose
native endian, because that's the same behavior as before.
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The way we're currently modeling mmio is too simplified. We assume that
every device has the same endianness as the target CPU. In reality,
most devices are little endian (all PCI and ISA ones I'm aware of). Some
are big endian (special system devices) and a very little fraction is
target native endian (fw_cfg).
So instead of assuming every device to be native endianness, let's move
to a model where the device tells us which endianness it's in.
That way we can compile the devices only once and get rid of all the ugly
swap will be done by the underlying layer.
For the same of readability, this patch only introduces the helper framework
but doesn't allow the registering code to set its endianness yet.
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Remove the debugging fprintf() slipped in via the following commit:
commit b2e0a138e7
Author: Michael S. Tsirkin <mst@redhat.com>
Date: Mon Nov 22 19:52:34 2010 +0200
migration: stable ram block ordering
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This makes ram block ordering under migration stable, ordered by offset.
This is especially useful for migration to exec, for debugging.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Jason Wang <jasowang@redhat.com>
fprintf_function uses format checking with GCC_FMT_ATTR.
It is declared in qemu-common.h and used in cpu-all.h
(which is included from cpu.h), so qemu-common.h must
be included earlier. Some redundant include statements
for standard include files were removed.
Fix also two format errors (ptrdiff_t needs %td).
Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
is_softmmu was removed with commit
d4c430a80f,
so remove it now from debug code, too.
Fix also the format specifier for paddr
in the same line of code.
Cc: Blue Swirl <blauwirbel@gmail.com>
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
vl.c has a Sun-specific hack to supply a prototype for madvise(),
but the call site has apparently moved to arch_init.c.
Haiku doesn't implement madvise() in favor of posix_madvise().
OpenBSD and Solaris 10 don't implement posix_madvise() but madvise().
MinGW implements neither.
Check for madvise() and posix_madvise() in configure and supply qemu_madvise()
as wrapper. Prefer madvise() over posix_madvise() due to flag availability.
Convert all callers to use qemu_madvise() and QEMU_MADV_*.
Note that on Solaris the warning is fixed by moving the madvise() prototype,
not by qemu_madvise() itself. It helps with porting though, and it simplifies
most call sites.
v7 -> v8:
* Some versions of MinGW have no sys/mman.h header. Reported by Blue Swirl.
v6 -> v7:
* Adopt madvise() rather than posix_madvise() semantics for returning errors.
* Use EINVAL in place of ENOTSUP.
v5 -> v6:
* Replace two leftover instances of POSIX_MADV_NORMAL with QEMU_MADV_INVALID.
Spotted by Blue Swirl.
v4 -> v5:
* Introduce QEMU_MADV_INVALID, suggested by Alexander Graf.
Note that this relies on -1 not being a valid advice value.
v3 -> v4:
* Eliminate #ifdefs at qemu_advise() call sites. Requested by Blue Swirl.
This will currently break the check in kvm-all.c by calling madvise() with
a supported flag, which will not fail. Ideas/patches welcome.
v2 -> v3:
* Reuse the *_MADV_* defines for QEMU_MADV_*. Suggested by Alexander Graf.
* Add configure check for madvise(), too.
Add defines to Makefile, not QEMU_CFLAGS.
Convert all callers, untested. Suggested by Blue Swirl.
* Keep Solaris' madvise() prototype around. Pointed out by Alexander Graf.
* Display configure check results.
v1 -> v2:
* Don't rely on posix_madvise() availability, add qemu_madvise().
Suggested by Blue Swirl.
Signed-off-by: Andreas Färber <afaerber@opensolaris.org>
Cc: Blue Swirl <blauwirbel@gmail.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
It is possible that subpage mmio is registered over existing memory
page. When this happens "memory" will have real memory address and not
index into io_mem array so next access to the page will generate
segfault. It is uncommon to have some part of a page to be accessed as
memory and some as mmio, but qemu shouldn't crash even when guest does
stupid things. So lets just pretend that the rest of the page is
unassigned if guest configure part of the memory page as mmio.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Since most of the code in qemu_ram_alloc() and
qemu_ram_alloc_from_ptr() are duplicated, let
qemu_ram_alloc_from_ptr() to switch by checking void *host, and change
qemu_ram_alloc() to a wrapper.
Signed-off-by: Yoshiaki Tamura <tamura.yoshiaki@lab.ntt.co.jp>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Provide a function to add an allocated region of memory to the qemu RAM.
This patch is copied from Marcelo's qemu_ram_map() in qemu-kvm and given the
clearer name qemu_ram_alloc_from_ptr().
Signed-off-by: Cam Macdonell <cam@cs.ualberta.ca>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Both values are only used in exec.c, so there is no need
to make them globally available.
Signed-off-by: Stefan Weil <weil@mail.berlios.de>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
With gcc 4.2.1-sjlj (mingw32-2) I get this warning:
/src/qemu/exec.c: In function 'qemu_ram_alloc':
/src/qemu/exec.c:2777: warning: 'offset' may be used uninitialized in this function
Fix by initializing the variable.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Now that we have a working qemu_ram_free() and the primary runtime
user of it has been updated, don't be lenient about duplicate id strings.
We also shouldn't need to create them ondemand at the target.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Now that we can support a ram_addr_t space with holes, we can implement
qemu_ram_free().
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
With these two pieces in place, we can start naming ramblocks. When
the device is present and it lives on a bus that provides a device
path, we concatenate the path and the provided name. Otherwise we
just use name. The resulting id string must be unique. For now we
assume an allocation for the same name and size is a device that has
been removed and reinserted and return the same block. This will go
away once qemu_ram_free() is implemented.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
These will be used to generate unique id strings for ramblocks. The name
field is required, the device pointer is optional as most callers don't
have a device. When there's no device or the device isn't a child of
a bus implementing BusInfo.get_dev_path, the name should be unique for
the platform.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
When available, we'd like to be able to access the DeviceState
when registering a savevm. For buses with a get_dev_path()
function, this will allow us to create more unique savevm
id strings.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
We currently need this either to allocate the next ram_addr_t for a
new block, or for total memory to be migrated. Both of which we can
calculate without need of this to keep us in a contiguous address space.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This patch avoids handling write watchpoints on read-only memory access.
It also breaks the searching loop for watchpoint once the setup for
handling watchpoint later is done.
Signed-off-by: Jun Koi <junkoi2004@gmail.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This makes the RAM block list easier to manipulate. Also incorporate
relevant variables into the RAMList struct.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Acked-by: Chris Wright <chrisw@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This allows the use of direct calls to the helpers,
and a direct branch back to the epilogue.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
This will allow backends to make intelligent choices about how
to implement GUEST_BASE.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Implement the "functions may be omitted with NULL pointer"
interface mentioned in the function block comment by transforming
NULL entries in the read/write arrays into calls to the
unassigned_mem family of functions.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
exec.c has a comment 'XXX: optimize' for lduw_phys/stw_phys,
so let's do it, along the lines of stl_phys.
The reason to address 16 bit accesses specifically is that virtio relies
on these accesses to be done atomically, using memset as we do now
breaks this assumption, which is reported to cause qemu with kvm
to read wrong index values under stress.
https://bugzilla.redhat.com/show_bug.cgi?id=525323
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
The usermode PAGE_RESERVED code is not required by the current mmap
implementation, and is already broken when guest_base != 0.
Unfortunately the bsd emulation still uses the old mmap implementation,
so we can't rip it out altogether.
Signed-off-by: Paul Brook <paul@codesourcery.com>
Greatly simplify the subpage implementation by not supporting
multiple devices at the same address at different widths. We
don't need full copies of mem_read/mem_write/opaque for each
address, only a single index back into the main io_mem_* arrays.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
V2 that uses endaddr = end-of-guest-address-space if !h2g_valid(endaddr)
after I found out that indeed works; and also disables the FreeBSD 6.x
/compat/linux/proc/self/maps fallback because it can return partial lines
if (at least I think that's the reason) the mappings change between
subsequent read() calls.
Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>
Acked-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Replaces direct phys_ram_dirty access with wrapper functions to prevent
direct access to the phys_ram_dirty bitmap.
Signed-off-by: Yoshiaki Tamura <tamura.yoshiaki@lab.ntt.co.jp>
Signed-off-by: OHMURA Kei <ohmura.kei@lab.ntt.co.jp>
Reviewed-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Historically the qemu tlb "addend" field was used for both RAM and IO accesses,
so needed to be able to hold both host addresses (unsigned long) and guest
physical addresses (target_phys_addr_t). However since the introduction of
the iotlb field it has only been used for RAM accesses.
This means we can change the type of addend to unsigned long, and remove
associated hacks in the big-endian TCG backends.
We can also remove the host dependence from target_phys_addr_t.
Signed-off-by: Paul Brook <paul@codesourcery.com>
On ia64, the default memory alignement is not enough for a code
alignement. To fix that, force static_code_gen_buffer alignment
to CODE_GEN_ALIGN.
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
When the host page size is bigger that the target one, unprotecting a
page should:
- mark all the target pages corresponding to the host page as writable
- invalidate all tb corresponding to the host page (and not the target
page)
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Use kinfo_getvmmap(3) on FeeBSD >= 7.x and /compat/linux/proc on older
FreeBSD. (kinfo_getvmmap is preferred since /compat/linux/proc is
usually only mounted on hosts also using the Linuxolator.)
This patch is a bit hacky because the includes needed for kinfo_getvmmap
conflict with other definitions in exec.c by default so I had to `trick
around' a little, but I built the result in FreeBSD 6.4-stable and
7.2-stable tbs and on 8-stable on the host so the hacks at least
should be stable. (If this is a problem maybe we could also move the
kinfo_getvmmap invocations into a seperate source file but that would
be more work...)
Signed-off-by: Juergen Lock <nox@jelal.kn-bremen.de>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Arrange various declarations so that also non-CPU code can access
them, adjust users.
Move CPU specific code to cpus.c.
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
QEMU uses a fixed page size for the CPU TLB. If the guest uses large
pages then we effectively split these into multiple smaller pages, and
populate the corresponding TLB entries on demand.
When the guest invalidates the TLB by virtual address we must invalidate
all entries covered by the large page. However the address used to
invalidate the entry may not be present in the QEMU TLB, so we do not
know which regions to clear.
Implementing a full vaiable size TLB is hard and slow, so just keep a
simple address/mask pair to record which addresses may have been mapped by
large pages. If the guest invalidates this region then flush the
whole TLB.
Signed-off-by: Paul Brook <paul@codesourcery.com>
The multi-level pagetable code fails to iterate ove all entries because
of the L2_BITS v.s. L2_SIZE thinko.
Signed-off-by: Paul Brook <paul@codesourcery.com>
Fixes warning:
CC sparc-bsd-user/exec.o
/src/qemu/exec.c: In function `page_check_range':
/src/qemu/exec.c:2375: warning: comparison is always true due to limited range of data type
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
The page tracking code in exec.c is used by both userspace and system
emulation. Userspace emulation uses it to track virtual pages, and
system emulation to track ram pages. Introduce a new type to hold this
kind of address.
Signed-off-by: Paul Brook <paul@codesourcery.com>
The addr < end comparison prevents iterating over the last
page in the guest address space; an iteration based on
length avoids this problem.
At the same time, assert that the given address is in the
guest address space.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Define L1_MAP_ADDR_SPACE_BITS to be either the virtual address size
(in user mode) or physical address size (in system mode), and use
that to size l1_map. This rewrites page_find_alloc, page_flush_tb,
and walk_memory_regions.
Use TARGET_PHYS_ADDR_SPACE_BITS for the physical memory map based
off of l1_phys_map. This rewrites page_phys_find_alloc and
phys_page_for_each.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Removes a set of ifdefs from exec.c.
Introduce TARGET_VIRT_ADDR_SPACE_BITS for all targets other
than Alpha. This will be used for page_find_alloc, which is
supposed to be using virtual addresses in the first place.
Signed-off-by: Richard Henderson <rth@twiddle.net>
This grand cleanup drops all reset and vmsave/load related
synchronization points in favor of four(!) generic hooks:
- cpu_synchronize_all_states in qemu_savevm_state_complete
(initial sync from kernel before vmsave)
- cpu_synchronize_all_post_init in qemu_loadvm_state
(writeback after vmload)
- cpu_synchronize_all_post_init in main after machine init
- cpu_synchronize_all_post_reset in qemu_system_reset
(writeback after system reset)
These writeback points + the existing one of VCPU exec after
cpu_synchronize_state map on three levels of writeback:
- KVM_PUT_RUNTIME_STATE (during runtime, other VCPUs continue to run)
- KVM_PUT_RESET_STATE (on synchronous system reset, all VCPUs stopped)
- KVM_PUT_FULL_STATE (on init or vmload, all VCPUs stopped as well)
This level is passed to the arch-specific VCPU state writing function
that will decide which concrete substates need to be written. That way,
no writer of load, save or reset functions that interact with in-kernel
KVM states will ever have to worry about synchronization again. That
also means that a lot of reasons for races, segfaults and deadlocks are
eliminated.
cpu_synchronize_state remains untouched, just as Anthony suggested. We
continue to need it before reading or writing of VCPU states that are
also tracked by in-kernel KVM subsystems.
Consequently, this patch removes many cpu_synchronize_state calls that
are now redundant, just like remaining explicit register syncs.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Port qemu-kvm's -mem-path and -mem-prealloc options. These are useful
for backing guest memory with huge pages via hugetlbfs.
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
CC: john cooper <john.cooper@redhat.com>
Userspace doesn't have physical memory, so cpu_physical_memory_rw
makes no sense. This is only used to implement cpu_memory_rw_debug, so
just implement that directly instead.
Signed-off-by: Paul Brook <paul@codesourcery.com>
Userspace emulation doesn't have a physical address space, so
l1_phys_map makes no sense. This code is never actually used, so don't
try and build it.
Signed-off-by: Paul Brook <paul@codesourcery.com>
remove direct kvm calls from exec.c, make
kvm use memory notifiers framework instead.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This adds notifiers for phys memory changes: a set of callbacks that
vhost can register and update kernel accordingly. Down the road, kvm
code can be switched to use these as well, instead of calling kvm code
directly from exec.c as is done now.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Qemu may hang in host_signal_handler after qemu has done a
seppuku with cpu_abort(). But at this stage we are not really
interested in target process coredump anymore, so unregister
host_signal_handler to die grafefully.
Signed-off-by: Riku Voipio <riku.voipio@nokia.com>
Richard Henderson