Extract RamDiscardManager and RamDiscardSource from system/memory.c into
dedicated a unit.
This reduces coupling and allows code that only needs the
RamDiscardManager interface to avoid pulling in all of memory.h
dependencies.
rust-sys bindings are no longer generated for RamDiscardSourceClass at
this point, thus we drop the unneeded InterfaceClass use.
Reviewed-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@kernel.org>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com>
Link: https://lore.kernel.org/r/20260604-rdm5-v5-2-5768e6a0943d@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Refactor the RamDiscardManager interface into two distinct components:
- RamDiscardSource: An interface that state providers (virtio-mem,
RamBlockAttributes) implement to provide discard state information
(granularity, populated/discarded ranges, replay callbacks).
- RamDiscardManager: A concrete QOM object that wraps a source, owns
the listener list, and handles listener registration/unregistration
and notifications.
This separation moves the listener management logic from individual
source implementations into the central RamDiscardManager, reducing
code duplication between virtio-mem and RamBlockAttributes.
The change prepares for future work where a RamDiscardManager could
aggregate multiple sources.
Note, the original virtio-mem code had conditions before discard:
if (vmem->size) {
rdl->notify_discard(rdl, rdl->section);
}
however, the new code calls discard unconditionally. This is considered
safe, since the populate/discard of sections are already asymmetrical
(unplug & unregister all listener section unconditionally).
Reviewed-by: Peter Xu <peterx@redhat.com>
Acked-by: David Hildenbrand <david@kernel.org>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@oss.qualcomm.com>
Link: https://lore.kernel.org/r/20260604-rdm5-v5-1-5768e6a0943d@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Some RAM device regions created with memory_region_init_ram_device_ptr()
are not intended to be P2P DMA targets.
The VFIO listener currently treats all RAM device regions as DMA
capable and attempts to map them into the IOMMU. For regions without
dma-buf backing this fails and prints warnings such as:
IOMMU_IOAS_MAP failed: Bad address, PCI BAR?
Introduce a MemoryRegion flag (ram_device_skip_iommu_map) to mark RAM
device regions that should not be IOMMU mapped, paired with
memory_region_skip_iommu_map() / memory_region_set_skip_iommu_map()
accessors. When the flag is set, the VFIO listener skips DMA mapping
for that region.
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Shameer Kolothum <skolothumtho@nvidia.com>
Message-id: 20260609112552.378999-21-skolothumtho@nvidia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Add SHMEM_MAP/UNMAP requests to vhost-user for dynamic management of
VIRTIO Shared Memory mappings.
This implementation introduces VirtioSharedMemoryMapping as a unified
QOM object that manages both the mapping metadata and MemoryRegion
lifecycle. This object provides reference-counted lifecycle management
with automatic cleanup of file descriptors and memory regions
through QOM finalization.
This request allows backends to dynamically map file descriptors into a
VIRTIO Shared Memory Region identified by their shmid. Maps are created
using memory_region_init_ram_from_fd() with configurable read/write
permissions, and the resulting MemoryRegions are added as subregions to
the shmem container region. The mapped memory is then advertised to the
guest VIRTIO drivers as a base address plus offset for reading and
writting according to the requested mmap flags.
The backend can unmap memory ranges within a given VIRTIO Shared Memory
Region to free resources. Upon receiving this message, the frontend
removes the MemoryRegion as a subregion and automatically unreferences
the VirtioSharedMemoryMapping object, triggering cleanup if no other
references exist.
Error handling has been improved to ensure consistent behavior across
handlers that manage their own vhost_user_send_resp() calls. Since
these handlers clear the VHOST_USER_NEED_REPLY_MASK flag, explicit
error checking ensures proper connection closure on failures,
maintaining the expected error flow.
Note the memory region commit for these operations needs to be delayed
until after we reply to the backend to avoid deadlocks. Otherwise,
the MemoryListener would send a VHOST_USER_SET_MEM_TABLE message
before the reply.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Albert Esteve <aesteve@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20260304165223.2166175-2-aesteve@redhat.com>
The PCI subsystem uses PCI_NO_PASID for requests-without-PASID, but VT-d
emulation uses IOMMU_NO_PASID internally (ecap.RPS==0). This leads to
conversion and checking code between PCI_NO_PASID and IOMMU_NO_PASID
throughout the implementation.
Refactor to use IOMMU PASID consistently within Intel IOMMU by storing
IOMMU PASID value in vtd_as->pasid. After this change, PCI_NO_PASID is
only used at three boundary points:
1. PCI_NO_PASID -> IOMMU_NO_PASID: Convert PCI PASID to IOMMU PASID in
vtd_find_add_as() and cache in vtd_as->pasid.
2. IOMMU_NO_PASID -> PCI_NO_PASID: Convert when notifying UNMAP events
via memory_region_notify_iommu() and returning IOMMUTLBEntry in
vtd_iommu_translate().
This eliminates conversion/checks in PASID table lookups, simplifies
invalidation logic with consistent PASID values, and improves code
readability. The PCI subsystem interface remains unchanged to maintain
compatibility with other IOMMU implementations that may not use PASID 0
for requests-without-PASID.
Suggested-by: Clement Mathieu--Drif <clement.mathieu--drif@bull.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20260527054658.1021096-11-zhenzhong.duan@intel.com>
Guard the native endian definition we want to remove by surrounding
it with TARGET_NOT_USING_LEGACY_NATIVE_ENDIAN_API #ifdef'ry.
Assign values to the enumerators so they stay unchanged.
Once a target gets cleaned we'll set the definition in the target
config, then the target won't be able to use the legacy API anymore.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20260109165058.59144-21-philmd@linaro.org>
We have 115 direct inclusions of "system/memory.h", and 91 headers
in include/ use it: hundreds of files have to process it.
However only one single header really uses the MemoryRegionCache
API: "hw/virtio/virtio-access.h". Split it out to a new header,
avoiding processing unused inlined functions hundreds of times.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-ID: <20260109165058.59144-6-philmd@linaro.org>
Commit 500131154d ("exec.c: Add new address_space_ld*/st*
functions") added a new API to fix a shortcoming of the
ld/st*_phys() API, which does blind bus access, not reporting
failure (and it also allow to provide transaction attributes).
Later commit 42874d3a8c ("Switch non-CPU callers from ld/st*_phys
to address_space_ld/st*") automatically converted the legacy uses
to the new API, not precising transaction attributes
(MEMTXATTRS_UNSPECIFIED) and ignoring the transation result (passing
NULL pointer as MemTxResult).
While this is a faithful replacement, without any logical change,
we later realized better is to not use MEMTXATTRS_UNSPECIFIED or
NULL MemTxResult, and adapt each call site on a pair basis, looking
at the device model datasheet to do the correct behavior (which is
unlikely to ignore transaction failures).
Since this is quite some work, we defer that to device model
maintainers. Meanwhile we introduce a definition, to allow a
target which removed all legacy API call to prohibit further
legacy API uses, named "TARGET_NOT_USING_LEGACY_LDST_PHYS_API".
Since all targets should be able to check this definition, we
take care to not poison it.
Suggested-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20251224151351.86733-6-philmd@linaro.org>
stl_phys_notdirty() was added in commit 8df1cd076c ("physical memory
access functions") as a (premature?) optimisation for the CODE path.
Meanwhile 20 years passed, we might never have understood / used it
properly; the code evolved and now the recommended way to access the
CODE path is via the cpu_ld/st_mmu*() API.
Remove both address_space_stl_notdirty() and stl_phys_notdirty()
leftovers.
Suggested-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20251224151351.86733-5-philmd@linaro.org>
Move RAMBlock functions out of ram_addr.h and cpu-common.h;
move memory API headers out of include/exec and into include/system.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Rename cpu_flush_icache_range() as address_space_flush_icache_range(),
passing an address space by argument. The single caller, rom_reset(),
already operates on an address space. Use it.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20251002084203.63899-7-philmd@linaro.org>
If an AddressSpace has been created in its own allocated
memory, cleaning it up requires first destroying the AS
and then freeing the memory. Doing this doesn't work:
address_space_destroy(as);
g_free_rcu(as, rcu);
because both address_space_destroy() and g_free_rcu()
try to use the same 'rcu' node in the AddressSpace struct
and the address_space_destroy hook gets overwritten.
Provide a new address_space_destroy_free() function which
will destroy the AS and then free the memory it uses, all
in one RCU callback.
(CC to stable because the next commit needs this function.)
Cc: qemu-stable@nongnu.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/20250929144228.1994037-3-peter.maydell@linaro.org
Signed-off-by: Peter Xu <peterx@redhat.com>
This will be necessary for devices implementing ATS.
We also define a new macro IOMMU_ACCESS_FLAG_FULL in addition to
IOMMU_ACCESS_FLAG to support more access flags.
IOMMU_ACCESS_FLAG is kept for convenience and backward compatibility.
Here are the flags added (defined by the PCIe 5 specification) :
- Execute Requested
- Privileged Mode Requested
- Global
- Untranslated Only
IOMMU_ACCESS_FLAG sets the additional flags to 0
Signed-off-by: Clement Mathieu--Drif <clement.mathieu--drif@eviden.com>
Message-Id: <20250628180226.133285-3-clement.mathieu--drif@eviden.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Modify memory_region_set_ram_discard_manager() to return -EBUSY if a
RamDiscardManager is already set in the MemoryRegion. The caller must
handle this failure, such as having virtio-mem undo its actions and fail
the realize() process. Opportunistically move the call earlier to avoid
complex error handling.
This change is beneficial when introducing a new RamDiscardManager
instance besides virtio-mem. After
ram_block_coordinated_discard_require(true) unlocks all
RamDiscardManager instances, only one instance is allowed to be set for
one MemoryRegion at present.
Suggested-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com>
Tested-by: Alexey Kardashevskiy <aik@amd.com>
Reviewed-by: Alexey Kardashevskiy <aik@amd.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Signed-off-by: Chenyi Qiang <chenyi.qiang@intel.com>
Link: https://lore.kernel.org/r/20250612082747.51539-3-chenyi.qiang@intel.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Convert the existing includes with
sed -i ,exec/memory.h,system/memory.h,g
Move the include within cpu-all.h into a !CONFIG_USER_ONLY block.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
