From 6c1b28e9e405da8cfb7296b9efa2b85650086784 Mon Sep 17 00:00:00 2001 From: David Hildenbrand Date: Thu, 22 Jun 2023 12:18:23 +0200 Subject: memory-device: Support empty memory devices Let's support empty memory devices -- memory devices that don't have a memory device region in the current configuration. hv-balloon with an optional memdev is the primary use case. Signed-off-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- hw/mem/memory-device.c | 43 +++++++++++++++++++++++++++++++++++++++--- include/hw/mem/memory-device.h | 7 ++++++- 2 files changed, 46 insertions(+), 4 deletions(-) diff --git a/hw/mem/memory-device.c b/hw/mem/memory-device.c index ae38f48..db702cc 100644 --- a/hw/mem/memory-device.c +++ b/hw/mem/memory-device.c @@ -20,6 +20,22 @@ #include "exec/address-spaces.h" #include "trace.h" +static bool memory_device_is_empty(const MemoryDeviceState *md) +{ + const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(md); + Error *local_err = NULL; + MemoryRegion *mr; + + /* dropping const here is fine as we don't touch the memory region */ + mr = mdc->get_memory_region((MemoryDeviceState *)md, &local_err); + if (local_err) { + /* Not empty, we'll report errors later when ontaining the MR again. */ + error_free(local_err); + return false; + } + return !mr; +} + static gint memory_device_addr_sort(gconstpointer a, gconstpointer b) { const MemoryDeviceState *md_a = MEMORY_DEVICE(a); @@ -249,6 +265,10 @@ static uint64_t memory_device_get_free_addr(MachineState *ms, uint64_t next_addr; Range tmp; + if (memory_device_is_empty(md)) { + continue; + } + range_init_nofail(&tmp, mdc->get_addr(md), memory_device_get_region_size(md, &error_abort)); @@ -292,6 +312,7 @@ MemoryDeviceInfoList *qmp_memory_device_list(void) const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(item->data); MemoryDeviceInfo *info = g_new0(MemoryDeviceInfo, 1); + /* Let's query infotmation even for empty memory devices. */ mdc->fill_device_info(md, info); QAPI_LIST_APPEND(tail, info); @@ -311,7 +332,7 @@ static int memory_device_plugged_size(Object *obj, void *opaque) const MemoryDeviceState *md = MEMORY_DEVICE(obj); const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(obj); - if (dev->realized) { + if (dev->realized && !memory_device_is_empty(md)) { *size += mdc->get_plugged_size(md, &error_abort); } } @@ -337,6 +358,11 @@ void memory_device_pre_plug(MemoryDeviceState *md, MachineState *ms, uint64_t addr, align = 0; MemoryRegion *mr; + /* We support empty memory devices even without device memory. */ + if (memory_device_is_empty(md)) { + return; + } + if (!ms->device_memory) { error_setg(errp, "the configuration is not prepared for memory devices" " (e.g., for memory hotplug), consider specifying the" @@ -380,10 +406,17 @@ out: void memory_device_plug(MemoryDeviceState *md, MachineState *ms) { const MemoryDeviceClass *mdc = MEMORY_DEVICE_GET_CLASS(md); - const unsigned int memslots = memory_device_get_memslots(md); - const uint64_t addr = mdc->get_addr(md); + unsigned int memslots; + uint64_t addr; MemoryRegion *mr; + if (memory_device_is_empty(md)) { + return; + } + + memslots = memory_device_get_memslots(md); + addr = mdc->get_addr(md); + /* * We expect that a previous call to memory_device_pre_plug() succeeded, so * it can't fail at this point. @@ -408,6 +441,10 @@ void memory_device_unplug(MemoryDeviceState *md, MachineState *ms) const unsigned int memslots = memory_device_get_memslots(md); MemoryRegion *mr; + if (memory_device_is_empty(md)) { + return; + } + /* * We expect that a previous call to memory_device_pre_plug() succeeded, so * it can't fail at this point. diff --git a/include/hw/mem/memory-device.h b/include/hw/mem/memory-device.h index 3354d6c..a1d62cc 100644 --- a/include/hw/mem/memory-device.h +++ b/include/hw/mem/memory-device.h @@ -38,6 +38,10 @@ typedef struct MemoryDeviceState MemoryDeviceState; * address in guest physical memory can either be specified explicitly * or get assigned automatically. * + * Some memory device might not own a memory region in certain device + * configurations. Such devices can logically get (un)plugged, however, + * empty memory devices are mostly ignored by the memory device code. + * * Conceptually, memory devices only span one memory region. If multiple * successive memory regions are used, a covering memory region has to * be provided. Scattered memory regions are not supported for single @@ -91,7 +95,8 @@ struct MemoryDeviceClass { uint64_t (*get_plugged_size)(const MemoryDeviceState *md, Error **errp); /* - * Return the memory region of the memory device. + * Return the memory region of the memory device. If the device is + * completely empty, returns NULL without an error. * * Called when (un)plugging the memory device, to (un)map the * memory region in guest physical memory, but also to detect the -- cgit v1.1 From 2d7f1081864790eb1000e6ef34e202dae66a03d2 Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Mon, 6 Nov 2023 13:53:59 +0100 Subject: Revert "hw/virtio/virtio-pmem: Replace impossible check by assertion" This reverts commit 5960f254dbb46f0c7a9f5f44bf4d27c19c34cb97 since the previous commit made this situation possible again. Signed-off-by: Maciej S. Szmigiero --- hw/virtio/virtio-pmem.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/hw/virtio/virtio-pmem.c b/hw/virtio/virtio-pmem.c index cc24812..c3512c2 100644 --- a/hw/virtio/virtio-pmem.c +++ b/hw/virtio/virtio-pmem.c @@ -147,7 +147,10 @@ static void virtio_pmem_fill_device_info(const VirtIOPMEM *pmem, static MemoryRegion *virtio_pmem_get_memory_region(VirtIOPMEM *pmem, Error **errp) { - assert(pmem->memdev); + if (!pmem->memdev) { + error_setg(errp, "'%s' property must be set", VIRTIO_PMEM_MEMDEV_PROP); + return NULL; + } return &pmem->memdev->mr; } -- cgit v1.1 From eb1b7c4bd4135648a96449d4607063e80692fd0c Mon Sep 17 00:00:00 2001 From: David Hildenbrand Date: Thu, 22 Jun 2023 12:49:19 +0200 Subject: memory-device: Drop size alignment check There is no strong requirement that the size has to be multiples of the requested alignment, let's drop it. This is a preparation for hv-baloon. Signed-off-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- hw/mem/memory-device.c | 6 ------ 1 file changed, 6 deletions(-) diff --git a/hw/mem/memory-device.c b/hw/mem/memory-device.c index db702cc..e0704b8 100644 --- a/hw/mem/memory-device.c +++ b/hw/mem/memory-device.c @@ -236,12 +236,6 @@ static uint64_t memory_device_get_free_addr(MachineState *ms, return 0; } - if (!QEMU_IS_ALIGNED(size, align)) { - error_setg(errp, "backend memory size must be multiple of 0x%" - PRIx64, align); - return 0; - } - if (hint) { if (range_init(&new, *hint, size) || !range_contains_range(&as, &new)) { error_setg(errp, "can't add memory device [0x%" PRIx64 ":0x%" PRIx64 -- cgit v1.1 From 4f80cd2f033e73bb65adf69bd337ae3052629fdf Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Mon, 12 Jun 2023 16:00:53 +0200 Subject: Add Hyper-V Dynamic Memory Protocol definitions This commit adds Hyper-V Dynamic Memory Protocol definitions, taken from hv_balloon Linux kernel driver, adapted to the QEMU coding style and definitions. Acked-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- include/hw/hyperv/dynmem-proto.h | 423 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 423 insertions(+) create mode 100644 include/hw/hyperv/dynmem-proto.h diff --git a/include/hw/hyperv/dynmem-proto.h b/include/hw/hyperv/dynmem-proto.h new file mode 100644 index 0000000..d0f9090 --- /dev/null +++ b/include/hw/hyperv/dynmem-proto.h @@ -0,0 +1,423 @@ +#ifndef HW_HYPERV_DYNMEM_PROTO_H +#define HW_HYPERV_DYNMEM_PROTO_H + +/* + * Hyper-V Dynamic Memory Protocol definitions + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * Based on drivers/hv/hv_balloon.c from Linux kernel: + * Copyright (c) 2012, Microsoft Corporation. + * + * Author: K. Y. Srinivasan + * + * This work is licensed under the terms of the GNU GPL, version 2. + * See the COPYING file in the top-level directory. + */ + +/* + * Protocol versions. The low word is the minor version, the high word the major + * version. + * + * History: + * Initial version 1.0 + * Changed to 0.1 on 2009/03/25 + * Changes to 0.2 on 2009/05/14 + * Changes to 0.3 on 2009/12/03 + * Changed to 1.0 on 2011/04/05 + * Changed to 2.0 on 2019/12/10 + */ + +#define DYNMEM_MAKE_VERSION(Major, Minor) ((uint32_t)(((Major) << 16) | (Minor))) +#define DYNMEM_MAJOR_VERSION(Version) ((uint32_t)(Version) >> 16) +#define DYNMEM_MINOR_VERSION(Version) ((uint32_t)(Version) & 0xff) + +enum { + DYNMEM_PROTOCOL_VERSION_1 = DYNMEM_MAKE_VERSION(0, 3), + DYNMEM_PROTOCOL_VERSION_2 = DYNMEM_MAKE_VERSION(1, 0), + DYNMEM_PROTOCOL_VERSION_3 = DYNMEM_MAKE_VERSION(2, 0), + + DYNMEM_PROTOCOL_VERSION_WIN7 = DYNMEM_PROTOCOL_VERSION_1, + DYNMEM_PROTOCOL_VERSION_WIN8 = DYNMEM_PROTOCOL_VERSION_2, + DYNMEM_PROTOCOL_VERSION_WIN10 = DYNMEM_PROTOCOL_VERSION_3, + + DYNMEM_PROTOCOL_VERSION_CURRENT = DYNMEM_PROTOCOL_VERSION_WIN10 +}; + + + +/* + * Message Types + */ + +enum dm_message_type { + /* + * Version 0.3 + */ + DM_ERROR = 0, + DM_VERSION_REQUEST = 1, + DM_VERSION_RESPONSE = 2, + DM_CAPABILITIES_REPORT = 3, + DM_CAPABILITIES_RESPONSE = 4, + DM_STATUS_REPORT = 5, + DM_BALLOON_REQUEST = 6, + DM_BALLOON_RESPONSE = 7, + DM_UNBALLOON_REQUEST = 8, + DM_UNBALLOON_RESPONSE = 9, + DM_MEM_HOT_ADD_REQUEST = 10, + DM_MEM_HOT_ADD_RESPONSE = 11, + DM_VERSION_03_MAX = 11, + /* + * Version 1.0. + */ + DM_INFO_MESSAGE = 12, + DM_VERSION_1_MAX = 12, + + /* + * Version 2.0 + */ + DM_MEM_HOT_REMOVE_REQUEST = 13, + DM_MEM_HOT_REMOVE_RESPONSE = 14 +}; + + +/* + * Structures defining the dynamic memory management + * protocol. + */ + +union dm_version { + struct { + uint16_t minor_version; + uint16_t major_version; + }; + uint32_t version; +} QEMU_PACKED; + + +union dm_caps { + struct { + uint64_t balloon:1; + uint64_t hot_add:1; + /* + * To support guests that may have alignment + * limitations on hot-add, the guest can specify + * its alignment requirements; a value of n + * represents an alignment of 2^n in mega bytes. + */ + uint64_t hot_add_alignment:4; + uint64_t hot_remove:1; + uint64_t reservedz:57; + } cap_bits; + uint64_t caps; +} QEMU_PACKED; + +union dm_mem_page_range { + struct { + /* + * The PFN number of the first page in the range. + * 40 bits is the architectural limit of a PFN + * number for AMD64. + */ + uint64_t start_page:40; + /* + * The number of pages in the range. + */ + uint64_t page_cnt:24; + } finfo; + uint64_t page_range; +} QEMU_PACKED; + + + +/* + * The header for all dynamic memory messages: + * + * type: Type of the message. + * size: Size of the message in bytes; including the header. + * trans_id: The guest is responsible for manufacturing this ID. + */ + +struct dm_header { + uint16_t type; + uint16_t size; + uint32_t trans_id; +} QEMU_PACKED; + +/* + * A generic message format for dynamic memory. + * Specific message formats are defined later in the file. + */ + +struct dm_message { + struct dm_header hdr; + uint8_t data[]; /* enclosed message */ +} QEMU_PACKED; + + +/* + * Specific message types supporting the dynamic memory protocol. + */ + +/* + * Version negotiation message. Sent from the guest to the host. + * The guest is free to try different versions until the host + * accepts the version. + * + * dm_version: The protocol version requested. + * is_last_attempt: If TRUE, this is the last version guest will request. + * reservedz: Reserved field, set to zero. + */ + +struct dm_version_request { + struct dm_header hdr; + union dm_version version; + uint32_t is_last_attempt:1; + uint32_t reservedz:31; +} QEMU_PACKED; + +/* + * Version response message; Host to Guest and indicates + * if the host has accepted the version sent by the guest. + * + * is_accepted: If TRUE, host has accepted the version and the guest + * should proceed to the next stage of the protocol. FALSE indicates that + * guest should re-try with a different version. + * + * reservedz: Reserved field, set to zero. + */ + +struct dm_version_response { + struct dm_header hdr; + uint64_t is_accepted:1; + uint64_t reservedz:63; +} QEMU_PACKED; + +/* + * Message reporting capabilities. This is sent from the guest to the + * host. + */ + +struct dm_capabilities { + struct dm_header hdr; + union dm_caps caps; + uint64_t min_page_cnt; + uint64_t max_page_number; +} QEMU_PACKED; + +/* + * Response to the capabilities message. This is sent from the host to the + * guest. This message notifies if the host has accepted the guest's + * capabilities. If the host has not accepted, the guest must shutdown + * the service. + * + * is_accepted: Indicates if the host has accepted guest's capabilities. + * reservedz: Must be 0. + */ + +struct dm_capabilities_resp_msg { + struct dm_header hdr; + uint64_t is_accepted:1; + uint64_t hot_remove:1; + uint64_t suppress_pressure_reports:1; + uint64_t reservedz:61; +} QEMU_PACKED; + +/* + * This message is used to report memory pressure from the guest. + * This message is not part of any transaction and there is no + * response to this message. + * + * num_avail: Available memory in pages. + * num_committed: Committed memory in pages. + * page_file_size: The accumulated size of all page files + * in the system in pages. + * zero_free: The nunber of zero and free pages. + * page_file_writes: The writes to the page file in pages. + * io_diff: An indicator of file cache efficiency or page file activity, + * calculated as File Cache Page Fault Count - Page Read Count. + * This value is in pages. + * + * Some of these metrics are Windows specific and fortunately + * the algorithm on the host side that computes the guest memory + * pressure only uses num_committed value. + */ + +struct dm_status { + struct dm_header hdr; + uint64_t num_avail; + uint64_t num_committed; + uint64_t page_file_size; + uint64_t zero_free; + uint32_t page_file_writes; + uint32_t io_diff; +} QEMU_PACKED; + + +/* + * Message to ask the guest to allocate memory - balloon up message. + * This message is sent from the host to the guest. The guest may not be + * able to allocate as much memory as requested. + * + * num_pages: number of pages to allocate. + */ + +struct dm_balloon { + struct dm_header hdr; + uint32_t num_pages; + uint32_t reservedz; +} QEMU_PACKED; + + +/* + * Balloon response message; this message is sent from the guest + * to the host in response to the balloon message. + * + * reservedz: Reserved; must be set to zero. + * more_pages: If FALSE, this is the last message of the transaction. + * if TRUE there will atleast one more message from the guest. + * + * range_count: The number of ranges in the range array. + * + * range_array: An array of page ranges returned to the host. + * + */ + +struct dm_balloon_response { + struct dm_header hdr; + uint32_t reservedz; + uint32_t more_pages:1; + uint32_t range_count:31; + union dm_mem_page_range range_array[]; +} QEMU_PACKED; + +/* + * Un-balloon message; this message is sent from the host + * to the guest to give guest more memory. + * + * more_pages: If FALSE, this is the last message of the transaction. + * if TRUE there will atleast one more message from the guest. + * + * reservedz: Reserved; must be set to zero. + * + * range_count: The number of ranges in the range array. + * + * range_array: An array of page ranges returned to the host. + * + */ + +struct dm_unballoon_request { + struct dm_header hdr; + uint32_t more_pages:1; + uint32_t reservedz:31; + uint32_t range_count; + union dm_mem_page_range range_array[]; +} QEMU_PACKED; + +/* + * Un-balloon response message; this message is sent from the guest + * to the host in response to an unballoon request. + * + */ + +struct dm_unballoon_response { + struct dm_header hdr; +} QEMU_PACKED; + + +/* + * Hot add request message. Message sent from the host to the guest. + * + * mem_range: Memory range to hot add. + * + */ + +struct dm_hot_add { + struct dm_header hdr; + union dm_mem_page_range range; +} QEMU_PACKED; + +/* + * Hot add response message. + * This message is sent by the guest to report the status of a hot add request. + * If page_count is less than the requested page count, then the host should + * assume all further hot add requests will fail, since this indicates that + * the guest has hit an upper physical memory barrier. + * + * Hot adds may also fail due to low resources; in this case, the guest must + * not complete this message until the hot add can succeed, and the host must + * not send a new hot add request until the response is sent. + * If VSC fails to hot add memory DYNMEM_NUMBER_OF_UNSUCCESSFUL_HOTADD_ATTEMPTS + * times it fails the request. + * + * + * page_count: number of pages that were successfully hot added. + * + * result: result of the operation 1: success, 0: failure. + * + */ + +struct dm_hot_add_response { + struct dm_header hdr; + uint32_t page_count; + uint32_t result; +} QEMU_PACKED; + +struct dm_hot_remove { + struct dm_header hdr; + uint32_t virtual_node; + uint32_t page_count; + uint32_t qos_flags; + uint32_t reservedZ; +} QEMU_PACKED; + +struct dm_hot_remove_response { + struct dm_header hdr; + uint32_t result; + uint32_t range_count; + uint64_t more_pages:1; + uint64_t reservedz:63; + union dm_mem_page_range range_array[]; +} QEMU_PACKED; + +#define DM_REMOVE_QOS_LARGE (1 << 0) +#define DM_REMOVE_QOS_LOCAL (1 << 1) +#define DM_REMOVE_QOS_MASK (0x3) + +/* + * Types of information sent from host to the guest. + */ + +enum dm_info_type { + INFO_TYPE_MAX_PAGE_CNT = 0, + MAX_INFO_TYPE +}; + + +/* + * Header for the information message. + */ + +struct dm_info_header { + enum dm_info_type type; + uint32_t data_size; + uint8_t data[]; +} QEMU_PACKED; + +/* + * This message is sent from the host to the guest to pass + * some relevant information (win8 addition). + * + * reserved: no used. + * info_size: size of the information blob. + * info: information blob. + */ + +struct dm_info_msg { + struct dm_header hdr; + uint32_t reserved; + uint32_t info_size; + uint8_t info[]; +}; + +#endif -- cgit v1.1 From 0d9e8c0b670b7856d36ed155d43548d2491230e7 Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Mon, 12 Jun 2023 16:00:54 +0200 Subject: Add Hyper-V Dynamic Memory Protocol driver (hv-balloon) base This driver is like virtio-balloon on steroids: it allows both changing the guest memory allocation via ballooning and (in the next patch) inserting pieces of extra RAM into it on demand from a provided memory backend. The actual resizing is done via ballooning interface (for example, via the "balloon" HMP command). This includes resizing the guest past its boot size - that is, hot-adding additional memory in granularity limited only by the guest alignment requirements, as provided by the next patch. In contrast with ACPI DIMM hotplug where one can only request to unplug a whole DIMM stick this driver allows removing memory from guest in single page (4k) units via ballooning. After a VM reboot the guest is back to its original (boot) size. In the future, the guest boot memory size might be changed on reboot instead, taking into account the effective size that VM had before that reboot (much like Hyper-V does). For performance reasons, the guest-released memory is tracked in a few range trees, as a series of (start, count) ranges. Each time a new page range is inserted into such tree its neighbors are checked as candidates for possible merging with it. Besides performance reasons, the Dynamic Memory protocol itself uses page ranges as the data structure in its messages, so relevant pages need to be merged into such ranges anyway. One has to be careful when tracking the guest-released pages, since the guest can maliciously report returning pages outside its current address space, which later clash with the address range of newly added memory. Similarly, the guest can report freeing the same page twice. The above design results in much better ballooning performance than when using virtio-balloon with the same guest: 230 GB / minute with this driver versus 70 GB / minute with virtio-balloon. During a ballooning operation most of time is spent waiting for the guest to come up with newly freed page ranges, processing the received ranges on the host side (in QEMU and KVM) is nearly instantaneous. The unballoon operation is also pretty much instantaneous: thanks to the merging of the ballooned out page ranges 200 GB of memory can be returned to the guest in about 1 second. With virtio-balloon this operation takes about 2.5 minutes. These tests were done against a Windows Server 2019 guest running on a Xeon E5-2699, after dirtying the whole memory inside guest before each balloon operation. Using a range tree instead of a bitmap to track the removed memory also means that the solution scales well with the guest size: even a 1 TB range takes just a few bytes of such metadata. Since the required GTree operations aren't present in every Glib version a check for them was added to the meson build script, together with new "--enable-hv-balloon" and "--disable-hv-balloon" configure arguments. If these GTree operations are missing in the system's Glib version this driver will be skipped during QEMU build. An optional "status-report=on" device parameter requests memory status events from the guest (typically sent every second), which allow the host to learn both the guest memory available and the guest memory in use counts. Following commits will add support for their external emission as "HV_BALLOON_STATUS_REPORT" QMP events. The driver is named hv-balloon since the Linux kernel client driver for the Dynamic Memory Protocol is named as such and to follow the naming pattern established by the virtio-balloon driver. The whole protocol runs over Hyper-V VMBus. The driver was tested against Windows Server 2012 R2, Windows Server 2016 and Windows Server 2019 guests and obeys the guest alignment requirements reported to the host via DM_CAPABILITIES_REPORT message. Acked-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- Kconfig.host | 3 + hw/hyperv/Kconfig | 10 + hw/hyperv/hv-balloon-internal.h | 33 + hw/hyperv/hv-balloon-page_range_tree.c | 228 +++++++ hw/hyperv/hv-balloon-page_range_tree.h | 118 ++++ hw/hyperv/hv-balloon.c | 1160 ++++++++++++++++++++++++++++++++ hw/hyperv/meson.build | 1 + hw/hyperv/trace-events | 13 + include/hw/hyperv/hv-balloon.h | 18 + meson.build | 28 +- meson_options.txt | 2 + scripts/meson-buildoptions.sh | 3 + 12 files changed, 1616 insertions(+), 1 deletion(-) create mode 100644 hw/hyperv/hv-balloon-internal.h create mode 100644 hw/hyperv/hv-balloon-page_range_tree.c create mode 100644 hw/hyperv/hv-balloon-page_range_tree.h create mode 100644 hw/hyperv/hv-balloon.c create mode 100644 include/hw/hyperv/hv-balloon.h diff --git a/Kconfig.host b/Kconfig.host index d763d89..2ee7157 100644 --- a/Kconfig.host +++ b/Kconfig.host @@ -46,3 +46,6 @@ config FUZZ config VFIO_USER_SERVER_ALLOWED bool imply VFIO_USER_SERVER + +config HV_BALLOON_POSSIBLE + bool diff --git a/hw/hyperv/Kconfig b/hw/hyperv/Kconfig index fcf6590..41dd827 100644 --- a/hw/hyperv/Kconfig +++ b/hw/hyperv/Kconfig @@ -16,3 +16,13 @@ config SYNDBG bool default y depends on VMBUS + +config HV_BALLOON_SUPPORTED + bool + +config HV_BALLOON + bool + default y + depends on VMBUS + depends on HV_BALLOON_POSSIBLE + depends on HV_BALLOON_SUPPORTED diff --git a/hw/hyperv/hv-balloon-internal.h b/hw/hyperv/hv-balloon-internal.h new file mode 100644 index 0000000..164c2e5 --- /dev/null +++ b/hw/hyperv/hv-balloon-internal.h @@ -0,0 +1,33 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#ifndef HW_HYPERV_HV_BALLOON_INTERNAL_H +#define HW_HYPERV_HV_BALLOON_INTERNAL_H + +#include "qemu/osdep.h" + +#define HV_BALLOON_PFN_SHIFT 12 +#define HV_BALLOON_PAGE_SIZE (1 << HV_BALLOON_PFN_SHIFT) + +#define SUM_OVERFLOW_U64(in1, in2) ((in1) > UINT64_MAX - (in2)) +#define SUM_SATURATE_U64(in1, in2) \ + ({ \ + uint64_t _in1 = (in1), _in2 = (in2); \ + uint64_t _result; \ + \ + if (!SUM_OVERFLOW_U64(_in1, _in2)) { \ + _result = _in1 + _in2; \ + } else { \ + _result = UINT64_MAX; \ + } \ + \ + _result; \ + }) + +#endif diff --git a/hw/hyperv/hv-balloon-page_range_tree.c b/hw/hyperv/hv-balloon-page_range_tree.c new file mode 100644 index 0000000..e178d8b --- /dev/null +++ b/hw/hyperv/hv-balloon-page_range_tree.c @@ -0,0 +1,228 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#include "hv-balloon-internal.h" +#include "hv-balloon-page_range_tree.h" + +/* + * temporarily avoid warnings about enhanced GTree API usage requiring a + * too recent Glib version until GLIB_VERSION_MAX_ALLOWED finally reaches + * the Glib version with this API + */ +#pragma GCC diagnostic ignored "-Wdeprecated-declarations" + +/* PageRangeTree */ +static gint page_range_tree_key_compare(gconstpointer leftp, + gconstpointer rightp, + gpointer user_data) +{ + const uint64_t *left = leftp, *right = rightp; + + if (*left < *right) { + return -1; + } else if (*left > *right) { + return 1; + } else { /* *left == *right */ + return 0; + } +} + +static GTreeNode *page_range_tree_insert_new(PageRangeTree tree, + uint64_t start, uint64_t count) +{ + uint64_t *key = g_malloc(sizeof(*key)); + PageRange *range = g_malloc(sizeof(*range)); + + assert(count > 0); + + *key = range->start = start; + range->count = count; + + return g_tree_insert_node(tree.t, key, range); +} + +void hvb_page_range_tree_insert(PageRangeTree tree, + uint64_t start, uint64_t count, + uint64_t *dupcount) +{ + GTreeNode *node; + bool joinable; + uint64_t intersection; + PageRange *range; + + assert(!SUM_OVERFLOW_U64(start, count)); + if (count == 0) { + return; + } + + node = g_tree_upper_bound(tree.t, &start); + if (node) { + node = g_tree_node_previous(node); + } else { + node = g_tree_node_last(tree.t); + } + + if (node) { + range = g_tree_node_value(node); + assert(range); + intersection = page_range_intersection_size(range, start, count); + joinable = page_range_joinable_right(range, start, count); + } + + if (!node || + (!intersection && !joinable)) { + /* + * !node case: the tree is empty or the very first node in the tree + * already has a higher key (the start of its range). + * the other case: there is a gap in the tree between the new range + * and the previous one. + * anyway, let's just insert the new range into the tree. + */ + node = page_range_tree_insert_new(tree, start, count); + assert(node); + range = g_tree_node_value(node); + assert(range); + } else { + /* + * the previous range in the tree either partially covers the new + * range or ends just at its beginning - extend it + */ + if (dupcount) { + *dupcount += intersection; + } + + count += start - range->start; + range->count = MAX(range->count, count); + } + + /* check next nodes for possible merging */ + for (node = g_tree_node_next(node); node; ) { + PageRange *rangecur; + + rangecur = g_tree_node_value(node); + assert(rangecur); + + intersection = page_range_intersection_size(rangecur, + range->start, range->count); + joinable = page_range_joinable_left(rangecur, + range->start, range->count); + if (!intersection && !joinable) { + /* the current node is disjoint */ + break; + } + + if (dupcount) { + *dupcount += intersection; + } + + count = rangecur->count + (rangecur->start - range->start); + range->count = MAX(range->count, count); + + /* the current node was merged in, remove it */ + start = rangecur->start; + node = g_tree_node_next(node); + /* no hinted removal in GTree... */ + g_tree_remove(tree.t, &start); + } +} + +bool hvb_page_range_tree_pop(PageRangeTree tree, PageRange *out, + uint64_t maxcount) +{ + GTreeNode *node; + PageRange *range; + + node = g_tree_node_last(tree.t); + if (!node) { + return false; + } + + range = g_tree_node_value(node); + assert(range); + + out->start = range->start; + + /* can't modify range->start as it is the node key */ + if (range->count > maxcount) { + out->start += range->count - maxcount; + out->count = maxcount; + range->count -= maxcount; + } else { + out->count = range->count; + /* no hinted removal in GTree... */ + g_tree_remove(tree.t, &out->start); + } + + return true; +} + +bool hvb_page_range_tree_intree_any(PageRangeTree tree, + uint64_t start, uint64_t count) +{ + GTreeNode *node; + + if (count == 0) { + return false; + } + + /* find the first node that can possibly intersect our range */ + node = g_tree_upper_bound(tree.t, &start); + if (node) { + /* + * a NULL node below means that the very first node in the tree + * already has a higher key (the start of its range). + */ + node = g_tree_node_previous(node); + } else { + /* a NULL node below means that the tree is empty */ + node = g_tree_node_last(tree.t); + } + /* node range start <= range start */ + + if (!node) { + /* node range start > range start */ + node = g_tree_node_first(tree.t); + } + + for ( ; node; node = g_tree_node_next(node)) { + PageRange *range = g_tree_node_value(node); + + assert(range); + /* + * if this node starts beyond or at the end of our range so does + * every next one + */ + if (range->start >= start + count) { + break; + } + + if (page_range_intersection_size(range, start, count) > 0) { + return true; + } + } + + return false; +} + +void hvb_page_range_tree_init(PageRangeTree *tree) +{ + tree->t = g_tree_new_full(page_range_tree_key_compare, NULL, + g_free, g_free); +} + +void hvb_page_range_tree_destroy(PageRangeTree *tree) +{ + /* g_tree_destroy() is not NULL-safe */ + if (!tree->t) { + return; + } + + g_tree_destroy(tree->t); + tree->t = NULL; +} diff --git a/hw/hyperv/hv-balloon-page_range_tree.h b/hw/hyperv/hv-balloon-page_range_tree.h new file mode 100644 index 0000000..07a9ae0 --- /dev/null +++ b/hw/hyperv/hv-balloon-page_range_tree.h @@ -0,0 +1,118 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#ifndef HW_HYPERV_HV_BALLOON_PAGE_RANGE_TREE_H +#define HW_HYPERV_HV_BALLOON_PAGE_RANGE_TREE_H + +#include "qemu/osdep.h" + +/* PageRange */ +typedef struct PageRange { + uint64_t start; + uint64_t count; +} PageRange; + +/* return just the part of range before (start) */ +static inline void page_range_part_before(const PageRange *range, + uint64_t start, PageRange *out) +{ + uint64_t endr = range->start + range->count; + uint64_t end = MIN(endr, start); + + out->start = range->start; + if (end > out->start) { + out->count = end - out->start; + } else { + out->count = 0; + } +} + +/* return just the part of range after (start, count) */ +static inline void page_range_part_after(const PageRange *range, + uint64_t start, uint64_t count, + PageRange *out) +{ + uint64_t end = range->start + range->count; + uint64_t ends = start + count; + + out->start = MAX(range->start, ends); + if (end > out->start) { + out->count = end - out->start; + } else { + out->count = 0; + } +} + +static inline void page_range_intersect(const PageRange *range, + uint64_t start, uint64_t count, + PageRange *out) +{ + uint64_t end1 = range->start + range->count; + uint64_t end2 = start + count; + uint64_t end = MIN(end1, end2); + + out->start = MAX(range->start, start); + out->count = out->start < end ? end - out->start : 0; +} + +static inline uint64_t page_range_intersection_size(const PageRange *range, + uint64_t start, uint64_t count) +{ + PageRange trange; + + page_range_intersect(range, start, count, &trange); + return trange.count; +} + +static inline bool page_range_joinable_left(const PageRange *range, + uint64_t start, uint64_t count) +{ + return start + count == range->start; +} + +static inline bool page_range_joinable_right(const PageRange *range, + uint64_t start, uint64_t count) +{ + return range->start + range->count == start; +} + +static inline bool page_range_joinable(const PageRange *range, + uint64_t start, uint64_t count) +{ + return page_range_joinable_left(range, start, count) || + page_range_joinable_right(range, start, count); +} + +/* PageRangeTree */ +/* type safety */ +typedef struct PageRangeTree { + GTree *t; +} PageRangeTree; + +static inline bool page_range_tree_is_empty(PageRangeTree tree) +{ + guint nnodes = g_tree_nnodes(tree.t); + + return nnodes == 0; +} + +void hvb_page_range_tree_init(PageRangeTree *tree); +void hvb_page_range_tree_destroy(PageRangeTree *tree); + +bool hvb_page_range_tree_intree_any(PageRangeTree tree, + uint64_t start, uint64_t count); + +bool hvb_page_range_tree_pop(PageRangeTree tree, PageRange *out, + uint64_t maxcount); + +void hvb_page_range_tree_insert(PageRangeTree tree, + uint64_t start, uint64_t count, + uint64_t *dupcount); + +#endif diff --git a/hw/hyperv/hv-balloon.c b/hw/hyperv/hv-balloon.c new file mode 100644 index 0000000..5f3674b --- /dev/null +++ b/hw/hyperv/hv-balloon.c @@ -0,0 +1,1160 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#include "hv-balloon-internal.h" + +#include "exec/address-spaces.h" +#include "exec/cpu-common.h" +#include "exec/ramblock.h" +#include "hw/boards.h" +#include "hw/hyperv/dynmem-proto.h" +#include "hw/hyperv/hv-balloon.h" +#include "hw/hyperv/vmbus.h" +#include "hw/mem/memory-device.h" +#include "hw/mem/pc-dimm.h" +#include "hw/qdev-core.h" +#include "hw/qdev-properties.h" +#include "monitor/qdev.h" +#include "qapi/error.h" +#include "qapi/qapi-commands-machine.h" +#include "qapi/qapi-events-machine.h" +#include "qapi/qapi-types-machine.h" +#include "qapi/qmp/qdict.h" +#include "qapi/visitor.h" +#include "qemu/error-report.h" +#include "qemu/module.h" +#include "qemu/units.h" +#include "qemu/timer.h" +#include "sysemu/balloon.h" +#include "sysemu/hostmem.h" +#include "sysemu/reset.h" +#include "hv-balloon-page_range_tree.h" +#include "trace.h" + +#define HV_BALLOON_GUID "525074DC-8985-46e2-8057-A307DC18A502" + +/* + * Some Windows versions (at least Server 2019) will crash with various + * error codes when receiving DM protocol requests (at least + * DM_MEM_HOT_ADD_REQUEST) immediately after boot. + * + * It looks like Hyper-V from Server 2016 uses a 50-second after-boot + * delay, probably to workaround this issue, so we'll use this value, too. + */ +#define HV_BALLOON_POST_INIT_WAIT (50 * 1000) + +#define HV_BALLOON_HA_CHUNK_SIZE (2 * GiB) +#define HV_BALLOON_HA_CHUNK_PAGES (HV_BALLOON_HA_CHUNK_SIZE / HV_BALLOON_PAGE_SIZE) + +#define HV_BALLOON_HR_CHUNK_PAGES 585728 +/* + * ^ that's the maximum number of pages + * that Windows returns in one hot remove response + * + * If the number requested is too high Windows will no longer honor + * these requests + */ + +struct HvBalloonClass { + VMBusDeviceClass parent_class; +} HvBalloonClass; + +typedef enum State { + /* not a real state */ + S_NO_CHANGE = 0, + + S_WAIT_RESET, + S_POST_RESET_CLOSED, + + /* init flow */ + S_VERSION, + S_CAPS, + S_POST_INIT_WAIT, + + S_IDLE, + + /* balloon op flow */ + S_BALLOON_POSTING, + S_BALLOON_RB_WAIT, + S_BALLOON_REPLY_WAIT, + + /* unballoon + hot add ops flow */ + S_UNBALLOON_POSTING, + S_UNBALLOON_RB_WAIT, + S_UNBALLOON_REPLY_WAIT, +} State; + +typedef struct StateDesc { + State state; + const char *desc; +} StateDesc; + +typedef struct HvBalloon { + VMBusDevice parent; + State state; + + union dm_version version; + union dm_caps caps; + + QEMUTimer post_init_timer; + + unsigned int trans_id; + + struct { + bool enabled; + bool received; + uint64_t committed; + uint64_t available; + } status_report; + + /* Guest target size */ + uint64_t target; + bool target_changed; + + /* Current (un)balloon */ + union { + uint64_t balloon_diff; + + struct { + uint64_t unballoon_diff; + }; + }; + + /* Nominal size of each memslot (the last one might be smaller) */ + uint64_t memslot_size; + + PageRangeTree removed_guest, removed_both; + + uint64_t removed_guest_ctr, removed_both_ctr; +} HvBalloon; + +OBJECT_DEFINE_TYPE_WITH_INTERFACES(HvBalloon, hv_balloon, HV_BALLOON, VMBUS_DEVICE, \ + { }) + +#define HV_BALLOON_SET_STATE(hvb, news) \ + do { \ + assert(news != S_NO_CHANGE); \ + hv_balloon_state_set(hvb, news, # news); \ + } while (0) + +#define HV_BALLOON_STATE_DESC_SET(stdesc, news) \ + _hv_balloon_state_desc_set(stdesc, news, # news) + +#define HV_BALLOON_STATE_DESC_INIT \ + { \ + .state = S_NO_CHANGE, \ + } + +typedef struct HvBalloonReq { + VMBusChanReq vmreq; +} HvBalloonReq; + +/* TODO: unify the code below with virtio-balloon and cache the value */ +static int build_dimm_list(Object *obj, void *opaque) +{ + GSList **list = opaque; + + if (object_dynamic_cast(obj, TYPE_PC_DIMM)) { + DeviceState *dev = DEVICE(obj); + if (dev->realized) { /* only realized DIMMs matter */ + *list = g_slist_prepend(*list, dev); + } + } + + object_child_foreach(obj, build_dimm_list, opaque); + return 0; +} + +static ram_addr_t get_current_ram_size(void) +{ + GSList *list = NULL, *item; + ram_addr_t size = current_machine->ram_size; + + build_dimm_list(qdev_get_machine(), &list); + for (item = list; item; item = g_slist_next(item)) { + Object *obj = OBJECT(item->data); + if (!strcmp(object_get_typename(obj), TYPE_PC_DIMM)) + size += object_property_get_int(obj, PC_DIMM_SIZE_PROP, + &error_abort); + } + g_slist_free(list); + + return size; +} + +/* total RAM includes memory currently removed from the guest */ +static uint64_t hv_balloon_total_ram(HvBalloon *balloon) +{ + ram_addr_t ram_size = get_current_ram_size(); + uint64_t ram_size_pages = ram_size >> HV_BALLOON_PFN_SHIFT; + + assert(ram_size_pages > 0); + + return ram_size_pages; +} + +/* + * calculating the total RAM size is a slow operation, + * avoid it as much as possible + */ +static uint64_t hv_balloon_total_removed_rs(HvBalloon *balloon, + uint64_t ram_size_pages) +{ + uint64_t total_removed; + + total_removed = SUM_SATURATE_U64(balloon->removed_guest_ctr, + balloon->removed_both_ctr); + + /* possible if guest returns pages outside actual RAM */ + if (total_removed > ram_size_pages) { + total_removed = ram_size_pages; + } + + return total_removed; +} + +/* Returns whether the state has actually changed */ +static bool hv_balloon_state_set(HvBalloon *balloon, + State newst, const char *newststr) +{ + if (newst == S_NO_CHANGE || balloon->state == newst) { + return false; + } + + balloon->state = newst; + trace_hv_balloon_state_change(newststr); + return true; +} + +static void _hv_balloon_state_desc_set(StateDesc *stdesc, + State newst, const char *newststr) +{ + /* state setting is only permitted on a freshly init desc */ + assert(stdesc->state == S_NO_CHANGE); + + assert(newst != S_NO_CHANGE); + + stdesc->state = newst; + stdesc->desc = newststr; +} + +static VMBusChannel *hv_balloon_get_channel_maybe(HvBalloon *balloon) +{ + return vmbus_device_channel(&balloon->parent, 0); +} + +static VMBusChannel *hv_balloon_get_channel(HvBalloon *balloon) +{ + VMBusChannel *chan; + + chan = hv_balloon_get_channel_maybe(balloon); + assert(chan != NULL); + return chan; +} + +static ssize_t hv_balloon_send_packet(VMBusChannel *chan, + struct dm_message *msg) +{ + int ret; + + ret = vmbus_channel_reserve(chan, 0, msg->hdr.size); + if (ret < 0) { + return ret; + } + + return vmbus_channel_send(chan, VMBUS_PACKET_DATA_INBAND, + NULL, 0, msg, msg->hdr.size, false, + msg->hdr.trans_id); +} + +static bool hv_balloon_unballoon_get_source(HvBalloon *balloon, + PageRangeTree *dtree, + uint64_t **dctr) +{ + if (g_tree_nnodes(balloon->removed_guest.t) > 0) { + *dtree = balloon->removed_guest; + *dctr = &balloon->removed_guest_ctr; + } else if (g_tree_nnodes(balloon->removed_both.t) > 0) { + *dtree = balloon->removed_both; + *dctr = &balloon->removed_both_ctr; + } else { + return false; + } + + return true; +} + +static void hv_balloon_unballoon_rb_wait(HvBalloon *balloon, StateDesc *stdesc) +{ + VMBusChannel *chan = hv_balloon_get_channel(balloon); + struct dm_unballoon_request *ur; + size_t ur_size = sizeof(*ur) + sizeof(ur->range_array[0]); + + assert(balloon->state == S_UNBALLOON_RB_WAIT); + + if (vmbus_channel_reserve(chan, 0, ur_size) < 0) { + return; + } + + HV_BALLOON_STATE_DESC_SET(stdesc, S_UNBALLOON_POSTING); +} + +static void hv_balloon_unballoon_posting(HvBalloon *balloon, StateDesc *stdesc) +{ + VMBusChannel *chan = hv_balloon_get_channel(balloon); + PageRangeTree dtree; + uint64_t *dctr; + struct dm_unballoon_request *ur; + size_t ur_size = sizeof(*ur) + sizeof(ur->range_array[0]); + PageRange range; + bool bret; + ssize_t ret; + + assert(balloon->state == S_UNBALLOON_POSTING); + assert(balloon->unballoon_diff > 0); + + if (!hv_balloon_unballoon_get_source(balloon, &dtree, &dctr)) { + error_report("trying to unballoon but nothing seems to be ballooned"); + /* + * there is little we can do as we might have already + * sent the guest a partial request we can't cancel + */ + return; + } + + assert(dtree.t); + assert(dctr); + + ur = alloca(ur_size); + memset(ur, 0, ur_size); + ur->hdr.type = DM_UNBALLOON_REQUEST; + ur->hdr.size = ur_size; + ur->hdr.trans_id = balloon->trans_id; + + bret = hvb_page_range_tree_pop(dtree, &range, MIN(balloon->unballoon_diff, + HV_BALLOON_HA_CHUNK_PAGES)); + assert(bret); + /* TODO: madvise? */ + + *dctr -= range.count; + balloon->unballoon_diff -= range.count; + + ur->range_count = 1; + ur->range_array[0].finfo.start_page = range.start; + ur->range_array[0].finfo.page_cnt = range.count; + ur->more_pages = balloon->unballoon_diff > 0; + + trace_hv_balloon_outgoing_unballoon(ur->hdr.trans_id, + range.count, range.start, + balloon->unballoon_diff); + + if (ur->more_pages) { + HV_BALLOON_STATE_DESC_SET(stdesc, S_UNBALLOON_RB_WAIT); + } else { + HV_BALLOON_STATE_DESC_SET(stdesc, S_UNBALLOON_REPLY_WAIT); + } + + ret = vmbus_channel_send(chan, VMBUS_PACKET_DATA_INBAND, + NULL, 0, ur, ur_size, false, + ur->hdr.trans_id); + if (ret <= 0) { + error_report("error %zd when posting unballoon msg, expect problems", + ret); + } +} + +static void hv_balloon_balloon_rb_wait(HvBalloon *balloon, StateDesc *stdesc) +{ + VMBusChannel *chan = hv_balloon_get_channel(balloon); + size_t bl_size = sizeof(struct dm_balloon); + + assert(balloon->state == S_BALLOON_RB_WAIT); + + if (vmbus_channel_reserve(chan, 0, bl_size) < 0) { + return; + } + + HV_BALLOON_STATE_DESC_SET(stdesc, S_BALLOON_POSTING); +} + +static void hv_balloon_balloon_posting(HvBalloon *balloon, StateDesc *stdesc) +{ + VMBusChannel *chan = hv_balloon_get_channel(balloon); + struct dm_balloon bl; + size_t bl_size = sizeof(bl); + ssize_t ret; + + assert(balloon->state == S_BALLOON_POSTING); + assert(balloon->balloon_diff > 0); + + memset(&bl, 0, sizeof(bl)); + bl.hdr.type = DM_BALLOON_REQUEST; + bl.hdr.size = bl_size; + bl.hdr.trans_id = balloon->trans_id; + bl.num_pages = MIN(balloon->balloon_diff, HV_BALLOON_HR_CHUNK_PAGES); + + trace_hv_balloon_outgoing_balloon(bl.hdr.trans_id, bl.num_pages, + balloon->balloon_diff); + + ret = vmbus_channel_send(chan, VMBUS_PACKET_DATA_INBAND, + NULL, 0, &bl, bl_size, false, + bl.hdr.trans_id); + if (ret <= 0) { + error_report("error %zd when posting balloon msg, expect problems", + ret); + } + + HV_BALLOON_STATE_DESC_SET(stdesc, S_BALLOON_REPLY_WAIT); +} + +static void hv_balloon_idle_state_process_target(HvBalloon *balloon, + StateDesc *stdesc) +{ + bool can_balloon = balloon->caps.cap_bits.balloon; + uint64_t ram_size_pages, total_removed; + + ram_size_pages = hv_balloon_total_ram(balloon); + total_removed = hv_balloon_total_removed_rs(balloon, ram_size_pages); + + /* + * we need to cache the values computed from the balloon target value when + * starting the adjustment procedure in case someone changes the target when + * the procedure is in progress + */ + if (balloon->target > ram_size_pages - total_removed) { + uint64_t target_diff = balloon->target - + (ram_size_pages - total_removed); + + balloon->unballoon_diff = MIN(target_diff, total_removed); + + if (balloon->unballoon_diff > 0) { + assert(can_balloon); + HV_BALLOON_STATE_DESC_SET(stdesc, S_UNBALLOON_RB_WAIT); + } + } else if (can_balloon && + balloon->target < ram_size_pages - total_removed) { + balloon->balloon_diff = ram_size_pages - total_removed - + balloon->target; + HV_BALLOON_STATE_DESC_SET(stdesc, S_BALLOON_RB_WAIT); + } +} + +static void hv_balloon_idle_state(HvBalloon *balloon, + StateDesc *stdesc) +{ + assert(balloon->state == S_IDLE); + + if (balloon->target_changed) { + balloon->target_changed = false; + hv_balloon_idle_state_process_target(balloon, stdesc); + return; + } +} + +static const struct { + void (*handler)(HvBalloon *balloon, StateDesc *stdesc); +} state_handlers[] = { + [S_IDLE].handler = hv_balloon_idle_state, + [S_BALLOON_POSTING].handler = hv_balloon_balloon_posting, + [S_BALLOON_RB_WAIT].handler = hv_balloon_balloon_rb_wait, + [S_UNBALLOON_POSTING].handler = hv_balloon_unballoon_posting, + [S_UNBALLOON_RB_WAIT].handler = hv_balloon_unballoon_rb_wait, +}; + +static void hv_balloon_handle_state(HvBalloon *balloon, StateDesc *stdesc) +{ + if (balloon->state >= ARRAY_SIZE(state_handlers) || + !state_handlers[balloon->state].handler) { + return; + } + + state_handlers[balloon->state].handler(balloon, stdesc); +} + +static void hv_balloon_remove_response_insert_range(PageRangeTree tree, + const PageRange *range, + uint64_t *ctr1, + uint64_t *ctr2, + uint64_t *ctr3) +{ + uint64_t dupcount, effcount; + + if (range->count == 0) { + return; + } + + dupcount = 0; + hvb_page_range_tree_insert(tree, range->start, range->count, &dupcount); + + assert(dupcount <= range->count); + effcount = range->count - dupcount; + + *ctr1 += effcount; + *ctr2 += effcount; + if (ctr3) { + *ctr3 += effcount; + } +} + +static void hv_balloon_remove_response_handle_range(HvBalloon *balloon, + PageRange *range, + bool both, + uint64_t *removedctr) +{ + PageRangeTree globaltree = + both ? balloon->removed_both : balloon->removed_guest; + uint64_t *globalctr = + both ? &balloon->removed_both_ctr : &balloon->removed_guest_ctr; + + trace_hv_balloon_remove_response(range->count, range->start, both); + + if (range->count > 0) { + hv_balloon_remove_response_insert_range(globaltree, range, + globalctr, removedctr, NULL); + trace_hv_balloon_remove_response_remainder(range->count, range->start, + both); + range->count = 0; + } +} + +static void hv_balloon_remove_response_handle_pages(HvBalloon *balloon, + PageRange *range, + uint64_t start, + uint64_t count, + bool both, + uint64_t *removedctr) +{ + assert(count > 0); + + /* + * if there is an existing range that the new range can't be joined to + * dump it into tree(s) + */ + if (range->count > 0 && !page_range_joinable(range, start, count)) { + hv_balloon_remove_response_handle_range(balloon, range, both, + removedctr); + } + + if (range->count == 0) { + range->start = start; + range->count = count; + } else if (page_range_joinable_left(range, start, count)) { + range->start = start; + range->count += count; + } else { /* page_range_joinable_right() */ + range->count += count; + } +} + +static gboolean hv_balloon_handle_remove_host_addr_node(gpointer key, + gpointer value, + gpointer data) +{ + PageRange *range = value; + uint64_t pageoff; + + for (pageoff = 0; pageoff < range->count; ) { + uint64_t addr_64 = (range->start + pageoff) * HV_BALLOON_PAGE_SIZE; + void *addr; + RAMBlock *rb; + ram_addr_t rb_offset; + size_t rb_page_size; + size_t discard_size; + + assert(addr_64 <= UINTPTR_MAX); + addr = (void *)((uintptr_t)addr_64); + rb = qemu_ram_block_from_host(addr, false, &rb_offset); + rb_page_size = qemu_ram_pagesize(rb); + + if (rb_page_size != HV_BALLOON_PAGE_SIZE) { + /* TODO: these should end in "removed_guest" */ + warn_report("guest reported removed page backed by unsupported page size %zu", + rb_page_size); + pageoff++; + continue; + } + + discard_size = MIN(range->count - pageoff, + (rb->max_length - rb_offset) / + HV_BALLOON_PAGE_SIZE); + discard_size = MAX(discard_size, 1); + + if (ram_block_discard_range(rb, rb_offset, discard_size * + HV_BALLOON_PAGE_SIZE) != 0) { + warn_report("guest reported removed page failed discard"); + } + + pageoff += discard_size; + } + + return false; +} + +static void hv_balloon_handle_remove_host_addr_tree(PageRangeTree tree) +{ + g_tree_foreach(tree.t, hv_balloon_handle_remove_host_addr_node, NULL); +} + +static int hv_balloon_handle_remove_section(PageRangeTree tree, + const MemoryRegionSection *section, + uint64_t count) +{ + void *addr = memory_region_get_ram_ptr(section->mr) + + section->offset_within_region; + uint64_t addr_page; + + assert(count > 0); + + if ((uintptr_t)addr % HV_BALLOON_PAGE_SIZE) { + warn_report("guest reported removed pages at an unaligned host addr %p", + addr); + return -EINVAL; + } + + addr_page = (uintptr_t)addr / HV_BALLOON_PAGE_SIZE; + hvb_page_range_tree_insert(tree, addr_page, count, NULL); + + return 0; +} + +static void hv_balloon_handle_remove_ranges(HvBalloon *balloon, + union dm_mem_page_range ranges[], + uint32_t count) +{ + uint64_t removedcnt; + PageRangeTree removed_host_addr; + PageRange range_guest, range_both; + + hvb_page_range_tree_init(&removed_host_addr); + range_guest.count = range_both.count = removedcnt = 0; + for (unsigned int ctr = 0; ctr < count; ctr++) { + union dm_mem_page_range *mr = &ranges[ctr]; + hwaddr pa; + MemoryRegionSection section; + + for (unsigned int offset = 0; offset < mr->finfo.page_cnt; ) { + int ret; + uint64_t pageno = mr->finfo.start_page + offset; + uint64_t pagecnt = 1; + + pa = (hwaddr)pageno << HV_BALLOON_PFN_SHIFT; + section = memory_region_find(get_system_memory(), pa, + (mr->finfo.page_cnt - offset) * + HV_BALLOON_PAGE_SIZE); + if (!section.mr) { + warn_report("guest reported removed page %"PRIu64" not found in RAM", + pageno); + ret = -EINVAL; + goto finish_page; + } + + pagecnt = int128_get64(section.size) / HV_BALLOON_PAGE_SIZE; + if (pagecnt <= 0) { + warn_report("guest reported removed page %"PRIu64" in a section smaller than page size", + pageno); + pagecnt = 1; /* skip the whole page */ + ret = -EINVAL; + goto finish_page; + } + + if (!memory_region_is_ram(section.mr) || + memory_region_is_rom(section.mr) || + memory_region_is_romd(section.mr)) { + warn_report("guest reported removed page %"PRIu64" in a section that is not an ordinary RAM", + pageno); + ret = -EINVAL; + goto finish_page; + } + + ret = hv_balloon_handle_remove_section(removed_host_addr, §ion, + pagecnt); + + finish_page: + if (ret == 0) { + hv_balloon_remove_response_handle_pages(balloon, + &range_both, + pageno, pagecnt, + true, &removedcnt); + } else { + hv_balloon_remove_response_handle_pages(balloon, + &range_guest, + pageno, pagecnt, + false, &removedcnt); + } + + if (section.mr) { + memory_region_unref(section.mr); + } + + offset += pagecnt; + } + } + + hv_balloon_remove_response_handle_range(balloon, &range_both, true, + &removedcnt); + hv_balloon_remove_response_handle_range(balloon, &range_guest, false, + &removedcnt); + + hv_balloon_handle_remove_host_addr_tree(removed_host_addr); + hvb_page_range_tree_destroy(&removed_host_addr); + + if (removedcnt > balloon->balloon_diff) { + warn_report("guest reported more pages removed than currently pending (%"PRIu64" vs %"PRIu64")", + removedcnt, balloon->balloon_diff); + balloon->balloon_diff = 0; + } else { + balloon->balloon_diff -= removedcnt; + } +} + +static bool hv_balloon_handle_msg_size(HvBalloonReq *req, size_t minsize, + const char *msgname) +{ + VMBusChanReq *vmreq = &req->vmreq; + uint32_t msglen = vmreq->msglen; + + if (msglen >= minsize) { + return true; + } + + warn_report("%s message too short (%u vs %zu), ignoring", msgname, + (unsigned int)msglen, minsize); + return false; +} + +static void hv_balloon_handle_version_request(HvBalloon *balloon, + HvBalloonReq *req, + StateDesc *stdesc) +{ + VMBusChanReq *vmreq = &req->vmreq; + struct dm_version_request *msgVr = vmreq->msg; + struct dm_version_response respVr; + + if (balloon->state != S_VERSION) { + warn_report("unexpected DM_VERSION_REQUEST in %d state", + balloon->state); + return; + } + + if (!hv_balloon_handle_msg_size(req, sizeof(*msgVr), + "DM_VERSION_REQUEST")) { + return; + } + + trace_hv_balloon_incoming_version(msgVr->version.major_version, + msgVr->version.minor_version); + + memset(&respVr, 0, sizeof(respVr)); + respVr.hdr.type = DM_VERSION_RESPONSE; + respVr.hdr.size = sizeof(respVr); + respVr.hdr.trans_id = msgVr->hdr.trans_id; + respVr.is_accepted = msgVr->version.version >= DYNMEM_PROTOCOL_VERSION_1 && + msgVr->version.version <= DYNMEM_PROTOCOL_VERSION_3; + + hv_balloon_send_packet(vmreq->chan, (struct dm_message *)&respVr); + + if (respVr.is_accepted) { + HV_BALLOON_STATE_DESC_SET(stdesc, S_CAPS); + } +} + +static void hv_balloon_handle_caps_report(HvBalloon *balloon, + HvBalloonReq *req, + StateDesc *stdesc) +{ + VMBusChanReq *vmreq = &req->vmreq; + struct dm_capabilities *msgCap = vmreq->msg; + struct dm_capabilities_resp_msg respCap; + + if (balloon->state != S_CAPS) { + warn_report("unexpected DM_CAPABILITIES_REPORT in %d state", + balloon->state); + return; + } + + if (!hv_balloon_handle_msg_size(req, sizeof(*msgCap), + "DM_CAPABILITIES_REPORT")) { + return; + } + + trace_hv_balloon_incoming_caps(msgCap->caps.caps); + balloon->caps = msgCap->caps; + + memset(&respCap, 0, sizeof(respCap)); + respCap.hdr.type = DM_CAPABILITIES_RESPONSE; + respCap.hdr.size = sizeof(respCap); + respCap.hdr.trans_id = msgCap->hdr.trans_id; + respCap.is_accepted = 1; + respCap.hot_remove = 1; + respCap.suppress_pressure_reports = !balloon->status_report.enabled; + hv_balloon_send_packet(vmreq->chan, (struct dm_message *)&respCap); + + timer_mod(&balloon->post_init_timer, + qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + + HV_BALLOON_POST_INIT_WAIT); + + HV_BALLOON_STATE_DESC_SET(stdesc, S_POST_INIT_WAIT); +} + +static void hv_balloon_handle_status_report(HvBalloon *balloon, + HvBalloonReq *req) +{ + VMBusChanReq *vmreq = &req->vmreq; + struct dm_status *msgStatus = vmreq->msg; + + if (!hv_balloon_handle_msg_size(req, sizeof(*msgStatus), + "DM_STATUS_REPORT")) { + return; + } + + if (!balloon->status_report.enabled) { + return; + } + + balloon->status_report.committed = msgStatus->num_committed; + balloon->status_report.committed *= HV_BALLOON_PAGE_SIZE; + balloon->status_report.available = msgStatus->num_avail; + balloon->status_report.available *= HV_BALLOON_PAGE_SIZE; + balloon->status_report.received = true; + + /* report event */ +} + +static void hv_balloon_handle_unballoon_response(HvBalloon *balloon, + HvBalloonReq *req, + StateDesc *stdesc) +{ + VMBusChanReq *vmreq = &req->vmreq; + struct dm_unballoon_response *msgUrR = vmreq->msg; + + if (balloon->state != S_UNBALLOON_REPLY_WAIT) { + warn_report("unexpected DM_UNBALLOON_RESPONSE in %d state", + balloon->state); + return; + } + + if (!hv_balloon_handle_msg_size(req, sizeof(*msgUrR), + "DM_UNBALLOON_RESPONSE")) + return; + + trace_hv_balloon_incoming_unballoon(msgUrR->hdr.trans_id); + + balloon->trans_id++; + + HV_BALLOON_STATE_DESC_SET(stdesc, S_IDLE); +} + +static void hv_balloon_handle_balloon_response(HvBalloon *balloon, + HvBalloonReq *req, + StateDesc *stdesc) +{ + VMBusChanReq *vmreq = &req->vmreq; + struct dm_balloon_response *msgBR = vmreq->msg; + + if (balloon->state != S_BALLOON_REPLY_WAIT) { + warn_report("unexpected DM_BALLOON_RESPONSE in %d state", + balloon->state); + return; + } + + if (!hv_balloon_handle_msg_size(req, sizeof(*msgBR), + "DM_BALLOON_RESPONSE")) + return; + + trace_hv_balloon_incoming_balloon(msgBR->hdr.trans_id, msgBR->range_count, + msgBR->more_pages); + + if (vmreq->msglen < sizeof(*msgBR) + + (uint64_t)sizeof(msgBR->range_array[0]) * msgBR->range_count) { + warn_report("DM_BALLOON_RESPONSE too short for the range count"); + return; + } + + if (msgBR->range_count == 0) { + /* The guest is already at its minimum size */ + balloon->balloon_diff = 0; + goto ret_end_trans; + } else { + hv_balloon_handle_remove_ranges(balloon, + msgBR->range_array, + msgBR->range_count); + } + + /* More responses expected? */ + if (msgBR->more_pages) { + return; + } + +ret_end_trans: + balloon->trans_id++; + + if (balloon->balloon_diff > 0) { + HV_BALLOON_STATE_DESC_SET(stdesc, S_BALLOON_RB_WAIT); + } else { + HV_BALLOON_STATE_DESC_SET(stdesc, S_IDLE); + } +} + +static void hv_balloon_handle_packet(HvBalloon *balloon, HvBalloonReq *req, + StateDesc *stdesc) +{ + VMBusChanReq *vmreq = &req->vmreq; + struct dm_message *msg = vmreq->msg; + + if (vmreq->msglen < sizeof(msg->hdr)) { + return; + } + + switch (msg->hdr.type) { + case DM_VERSION_REQUEST: + hv_balloon_handle_version_request(balloon, req, stdesc); + break; + + case DM_CAPABILITIES_REPORT: + hv_balloon_handle_caps_report(balloon, req, stdesc); + break; + + case DM_STATUS_REPORT: + hv_balloon_handle_status_report(balloon, req); + break; + + case DM_UNBALLOON_RESPONSE: + hv_balloon_handle_unballoon_response(balloon, req, stdesc); + break; + + case DM_BALLOON_RESPONSE: + hv_balloon_handle_balloon_response(balloon, req, stdesc); + break; + + default: + warn_report("unknown DM message %u", msg->hdr.type); + break; + } +} + +static bool hv_balloon_recv_channel(HvBalloon *balloon, StateDesc *stdesc) +{ + VMBusChannel *chan; + HvBalloonReq *req; + + if (balloon->state == S_WAIT_RESET || + balloon->state == S_POST_RESET_CLOSED) { + return false; + } + + chan = hv_balloon_get_channel(balloon); + if (vmbus_channel_recv_start(chan)) { + return false; + } + + while ((req = vmbus_channel_recv_peek(chan, sizeof(*req)))) { + hv_balloon_handle_packet(balloon, req, stdesc); + vmbus_free_req(req); + vmbus_channel_recv_pop(chan); + + if (stdesc->state != S_NO_CHANGE) { + break; + } + } + + return vmbus_channel_recv_done(chan) > 0; +} + +/* old state handler -> new state transition (potential) */ +static bool hv_balloon_event_loop_state(HvBalloon *balloon) +{ + StateDesc state_new = HV_BALLOON_STATE_DESC_INIT; + + hv_balloon_handle_state(balloon, &state_new); + return hv_balloon_state_set(balloon, state_new.state, state_new.desc); +} + +/* VMBus message -> new state transition (potential) */ +static bool hv_balloon_event_loop_recv(HvBalloon *balloon) +{ + StateDesc state_new = HV_BALLOON_STATE_DESC_INIT; + bool any_recv, state_changed; + + any_recv = hv_balloon_recv_channel(balloon, &state_new); + state_changed = hv_balloon_state_set(balloon, + state_new.state, state_new.desc); + + return state_changed || any_recv; +} + +static void hv_balloon_event_loop(HvBalloon *balloon) +{ + bool state_repeat, recv_repeat; + + do { + state_repeat = hv_balloon_event_loop_state(balloon); + recv_repeat = hv_balloon_event_loop_recv(balloon); + } while (state_repeat || recv_repeat); +} + +static void hv_balloon_vmdev_chan_notify(VMBusChannel *chan) +{ + HvBalloon *balloon = HV_BALLOON(vmbus_channel_device(chan)); + + hv_balloon_event_loop(balloon); +} + +static void hv_balloon_stat(void *opaque, BalloonInfo *info) +{ + HvBalloon *balloon = opaque; + info->actual = (hv_balloon_total_ram(balloon) - balloon->removed_both_ctr) + << HV_BALLOON_PFN_SHIFT; +} + +static void hv_balloon_to_target(void *opaque, ram_addr_t target) +{ + HvBalloon *balloon = opaque; + uint64_t target_pages = target >> HV_BALLOON_PFN_SHIFT; + + if (!target_pages) { + return; + } + + /* + * always set target_changed, even with unchanged target, as the user + * might be asking us to try again reaching it + */ + balloon->target = target_pages; + balloon->target_changed = true; + + hv_balloon_event_loop(balloon); +} + +static int hv_balloon_vmdev_open_channel(VMBusChannel *chan) +{ + HvBalloon *balloon = HV_BALLOON(vmbus_channel_device(chan)); + + if (balloon->state != S_POST_RESET_CLOSED) { + warn_report("guest trying to open a DM channel in invalid %d state", + balloon->state); + return -EINVAL; + } + + HV_BALLOON_SET_STATE(balloon, S_VERSION); + hv_balloon_event_loop(balloon); + + return 0; +} + +static void hv_balloon_vmdev_close_channel(VMBusChannel *chan) +{ + HvBalloon *balloon = HV_BALLOON(vmbus_channel_device(chan)); + + timer_del(&balloon->post_init_timer); + + /* Don't report stale data */ + balloon->status_report.received = false; + + HV_BALLOON_SET_STATE(balloon, S_WAIT_RESET); + hv_balloon_event_loop(balloon); +} + +static void hv_balloon_post_init_timer(void *opaque) +{ + HvBalloon *balloon = opaque; + + if (balloon->state != S_POST_INIT_WAIT) { + return; + } + + HV_BALLOON_SET_STATE(balloon, S_IDLE); + hv_balloon_event_loop(balloon); +} + +static void hv_balloon_vmdev_realize(VMBusDevice *vdev, Error **errp) +{ + ERRP_GUARD(); + HvBalloon *balloon = HV_BALLOON(vdev); + int ret; + + balloon->state = S_WAIT_RESET; + + ret = qemu_add_balloon_handler(hv_balloon_to_target, hv_balloon_stat, + balloon); + if (ret < 0) { + /* This also protects against having multiple hv-balloon instances */ + error_setg(errp, "Only one balloon device is supported"); + return; + } + + timer_init_ms(&balloon->post_init_timer, QEMU_CLOCK_VIRTUAL, + hv_balloon_post_init_timer, balloon); +} + +/* + * VMBus device reset has to be implemented in case the guest decides to + * disconnect and reconnect to the VMBus without rebooting the whole system. + */ +static void hv_balloon_vmdev_reset(VMBusDevice *vdev) +{ + HvBalloon *balloon = HV_BALLOON(vdev); + + if (balloon->state == S_POST_RESET_CLOSED) { + return; + } + + hvb_page_range_tree_destroy(&balloon->removed_guest); + hvb_page_range_tree_destroy(&balloon->removed_both); + hvb_page_range_tree_init(&balloon->removed_guest); + hvb_page_range_tree_init(&balloon->removed_both); + + balloon->trans_id = 0; + balloon->removed_guest_ctr = 0; + balloon->removed_both_ctr = 0; + + HV_BALLOON_SET_STATE(balloon, S_POST_RESET_CLOSED); + hv_balloon_event_loop(balloon); +} + +static void hv_balloon_vmdev_unrealize(VMBusDevice *vdev) +{ + HvBalloon *balloon = HV_BALLOON(vdev); + + qemu_remove_balloon_handler(balloon); + + hvb_page_range_tree_destroy(&balloon->removed_guest); + hvb_page_range_tree_destroy(&balloon->removed_both); +} + +static void hv_balloon_init(Object *obj) +{ +} + +static void hv_balloon_finalize(Object *obj) +{ +} + +static Property hv_balloon_properties[] = { + DEFINE_PROP_BOOL("status-report", HvBalloon, + status_report.enabled, false), + + DEFINE_PROP_END_OF_LIST(), +}; + +static void hv_balloon_class_init(ObjectClass *klass, void *data) +{ + DeviceClass *dc = DEVICE_CLASS(klass); + VMBusDeviceClass *vdc = VMBUS_DEVICE_CLASS(klass); + + device_class_set_props(dc, hv_balloon_properties); + qemu_uuid_parse(HV_BALLOON_GUID, &vdc->classid); + set_bit(DEVICE_CATEGORY_MISC, dc->categories); + + vdc->vmdev_realize = hv_balloon_vmdev_realize; + vdc->vmdev_unrealize = hv_balloon_vmdev_unrealize; + vdc->vmdev_reset = hv_balloon_vmdev_reset; + vdc->open_channel = hv_balloon_vmdev_open_channel; + vdc->close_channel = hv_balloon_vmdev_close_channel; + vdc->chan_notify_cb = hv_balloon_vmdev_chan_notify; +} diff --git a/hw/hyperv/meson.build b/hw/hyperv/meson.build index b43f119..b5be1cb 100644 --- a/hw/hyperv/meson.build +++ b/hw/hyperv/meson.build @@ -2,3 +2,4 @@ specific_ss.add(when: 'CONFIG_HYPERV', if_true: files('hyperv.c')) specific_ss.add(when: 'CONFIG_HYPERV_TESTDEV', if_true: files('hyperv_testdev.c')) specific_ss.add(when: 'CONFIG_VMBUS', if_true: files('vmbus.c')) specific_ss.add(when: 'CONFIG_SYNDBG', if_true: files('syndbg.c')) +specific_ss.add(when: 'CONFIG_HV_BALLOON', if_true: files('hv-balloon.c', 'hv-balloon-page_range_tree.c')) diff --git a/hw/hyperv/trace-events b/hw/hyperv/trace-events index b4c35ca..86c724b 100644 --- a/hw/hyperv/trace-events +++ b/hw/hyperv/trace-events @@ -16,3 +16,16 @@ vmbus_gpadl_torndown(uint32_t gpadl_id) "gpadl #%d" vmbus_open_channel(uint32_t chan_id, uint32_t gpadl_id, uint32_t target_vp) "channel #%d gpadl #%d target vp %d" vmbus_channel_open(uint32_t chan_id, uint32_t status) "channel #%d status %d" vmbus_close_channel(uint32_t chan_id) "channel #%d" + +# hv-balloon +hv_balloon_state_change(const char *tostr) "-> %s" +hv_balloon_incoming_version(uint16_t major, uint16_t minor) "incoming proto version %u.%u" +hv_balloon_incoming_caps(uint32_t caps) "incoming caps 0x%x" +hv_balloon_outgoing_unballoon(uint32_t trans_id, uint64_t count, uint64_t start, uint64_t rempages) "posting unballoon %"PRIu32" for %"PRIu64" @ 0x%"PRIx64", remaining %"PRIu64 +hv_balloon_incoming_unballoon(uint32_t trans_id) "incoming unballoon response %"PRIu32 +hv_balloon_outgoing_balloon(uint32_t trans_id, uint64_t count, uint64_t rempages) "posting balloon %"PRIu32" for %"PRIu64", remaining %"PRIu64 +hv_balloon_incoming_balloon(uint32_t trans_id, uint32_t range_count, uint32_t more_pages) "incoming balloon response %"PRIu32", ranges %"PRIu32", more %"PRIu32 +hv_balloon_remove_response(uint64_t count, uint64_t start, unsigned int both) "processing remove response range %"PRIu64" @ 0x%"PRIx64", both %u" +hv_balloon_remove_response_hole(uint64_t counthole, uint64_t starthole, uint64_t countrange, uint64_t startrange, uint64_t starthpr, unsigned int both) "response range hole %"PRIu64" @ 0x%"PRIx64" from range %"PRIu64" @ 0x%"PRIx64", before our start 0x%"PRIx64", both %u" +hv_balloon_remove_response_common(uint64_t countcommon, uint64_t startcommon, uint64_t countrange, uint64_t startrange, uint64_t counthpr, uint64_t starthpr, uint64_t removed, unsigned int both) "response common range %"PRIu64" @ 0x%"PRIx64" from range %"PRIu64" @ 0x%"PRIx64" with our %"PRIu64" @ 0x%"PRIx64", removed %"PRIu64", both %u" +hv_balloon_remove_response_remainder(uint64_t count, uint64_t start, unsigned int both) "remove response remaining range %"PRIu64" @ 0x%"PRIx64", both %u" diff --git a/include/hw/hyperv/hv-balloon.h b/include/hw/hyperv/hv-balloon.h new file mode 100644 index 0000000..c1efe70 --- /dev/null +++ b/include/hw/hyperv/hv-balloon.h @@ -0,0 +1,18 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#ifndef HW_HV_BALLOON_H +#define HW_HV_BALLOON_H + +#include "qom/object.h" + +#define TYPE_HV_BALLOON "hv-balloon" +OBJECT_DECLARE_SIMPLE_TYPE(HvBalloon, HV_BALLOON) + +#endif diff --git a/meson.build b/meson.build index dcef8b1..51a5107 100644 --- a/meson.build +++ b/meson.build @@ -1323,6 +1323,30 @@ if not get_option('glusterfs').auto() or have_block endif endif +hv_balloon = false +if get_option('hv_balloon').allowed() and have_system + if cc.links(''' + #include + #include + int main(void) { + GTree *tree; + + tree = g_tree_new((GCompareFunc)strcmp); + (void)g_tree_node_first(tree); + g_tree_destroy(tree); + return 0; + } + ''', dependencies: glib) + hv_balloon = true + else + if get_option('hv_balloon').enabled() + error('could not enable hv-balloon, update your glib') + else + warning('could not find glib support for hv-balloon, disabling') + endif + endif +endif + libssh = not_found if not get_option('libssh').auto() or have_block libssh = dependency('libssh', version: '>=0.8.7', @@ -2855,7 +2879,8 @@ host_kconfig = \ (targetos == 'linux' ? ['CONFIG_LINUX=y'] : []) + \ (have_pvrdma ? ['CONFIG_PVRDMA=y'] : []) + \ (multiprocess_allowed ? ['CONFIG_MULTIPROCESS_ALLOWED=y'] : []) + \ - (vfio_user_server_allowed ? ['CONFIG_VFIO_USER_SERVER_ALLOWED=y'] : []) + (vfio_user_server_allowed ? ['CONFIG_VFIO_USER_SERVER_ALLOWED=y'] : []) + \ + (hv_balloon ? ['CONFIG_HV_BALLOON_POSSIBLE=y'] : []) ignored = [ 'TARGET_XML_FILES', 'TARGET_ABI_DIR', 'TARGET_ARCH' ] @@ -4321,6 +4346,7 @@ if targetos == 'windows' endif summary_info += {'seccomp support': seccomp} summary_info += {'GlusterFS support': glusterfs} +summary_info += {'hv-balloon support': hv_balloon} summary_info += {'TPM support': have_tpm} summary_info += {'libssh support': libssh} summary_info += {'lzo support': lzo} diff --git a/meson_options.txt b/meson_options.txt index 3c7398f..5c212fc 100644 --- a/meson_options.txt +++ b/meson_options.txt @@ -150,6 +150,8 @@ option('gio', type : 'feature', value : 'auto', description: 'use libgio for D-Bus support') option('glusterfs', type : 'feature', value : 'auto', description: 'Glusterfs block device driver') +option('hv_balloon', type : 'feature', value : 'auto', + description: 'hv-balloon driver (requires Glib 2.68+ GTree API)') option('libdw', type : 'feature', value : 'auto', description: 'debuginfo support') option('libiscsi', type : 'feature', value : 'auto', diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh index 7ca4b77..e9d6d39 100644 --- a/scripts/meson-buildoptions.sh +++ b/scripts/meson-buildoptions.sh @@ -123,6 +123,7 @@ meson_options_help() { printf "%s\n" ' gtk-clipboard clipboard support for the gtk UI (EXPERIMENTAL, MAY HANG)' printf "%s\n" ' guest-agent Build QEMU Guest Agent' printf "%s\n" ' guest-agent-msi Build MSI package for the QEMU Guest Agent' + printf "%s\n" ' hv-balloon hv-balloon driver (requires Glib 2.68+ GTree API)' printf "%s\n" ' hvf HVF acceleration support' printf "%s\n" ' iconv Font glyph conversion support' printf "%s\n" ' jack JACK sound support' @@ -333,6 +334,8 @@ _meson_option_parse() { --disable-guest-agent-msi) printf "%s" -Dguest_agent_msi=disabled ;; --enable-hexagon-idef-parser) printf "%s" -Dhexagon_idef_parser=true ;; --disable-hexagon-idef-parser) printf "%s" -Dhexagon_idef_parser=false ;; + --enable-hv-balloon) printf "%s" -Dhv_balloon=enabled ;; + --disable-hv-balloon) printf "%s" -Dhv_balloon=disabled ;; --enable-hvf) printf "%s" -Dhvf=enabled ;; --disable-hvf) printf "%s" -Dhvf=disabled ;; --iasl=*) quote_sh "-Diasl=$2" ;; -- cgit v1.1 From 99a4706ae81efa51b21871af643626730a6719d4 Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Sun, 22 Oct 2023 17:20:20 +0200 Subject: Add Hyper-V Dynamic Memory Protocol driver (hv-balloon) hot-add support One of advantages of using this protocol over ACPI-based PC DIMM hotplug is that it allows hot-adding memory in much smaller granularity because the ACPI DIMM slot limit does not apply. In order to enable this functionality a new memory backend needs to be created and provided to the driver via the "memdev" parameter. This can be achieved by, for example, adding "-object memory-backend-ram,id=mem1,size=32G" to the QEMU command line and then instantiating the driver with "memdev=mem1" parameter. The device will try to use multiple memslots to cover the memory backend in order to reduce the size of metadata for the not-yet-hot-added part of the memory backend. Co-developed-by: David Hildenbrand Acked-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- hw/hyperv/hv-balloon-our_range_memslots.c | 201 +++++++++++ hw/hyperv/hv-balloon-our_range_memslots.h | 110 ++++++ hw/hyperv/hv-balloon.c | 566 +++++++++++++++++++++++++++++- hw/hyperv/meson.build | 2 +- hw/hyperv/trace-events | 5 + 5 files changed, 878 insertions(+), 6 deletions(-) create mode 100644 hw/hyperv/hv-balloon-our_range_memslots.c create mode 100644 hw/hyperv/hv-balloon-our_range_memslots.h diff --git a/hw/hyperv/hv-balloon-our_range_memslots.c b/hw/hyperv/hv-balloon-our_range_memslots.c new file mode 100644 index 0000000..99bae87 --- /dev/null +++ b/hw/hyperv/hv-balloon-our_range_memslots.c @@ -0,0 +1,201 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#include "hv-balloon-internal.h" +#include "hv-balloon-our_range_memslots.h" +#include "trace.h" + +/* OurRange */ +static void our_range_init(OurRange *our_range, uint64_t start, uint64_t count) +{ + assert(count <= UINT64_MAX - start); + our_range->range.start = start; + our_range->range.count = count; + + hvb_page_range_tree_init(&our_range->removed_guest); + hvb_page_range_tree_init(&our_range->removed_both); + + /* mark the whole range as unused but for potential use */ + our_range->added = 0; + our_range->unusable_tail = 0; +} + +static void our_range_destroy(OurRange *our_range) +{ + hvb_page_range_tree_destroy(&our_range->removed_guest); + hvb_page_range_tree_destroy(&our_range->removed_both); +} + +void hvb_our_range_clear_removed_trees(OurRange *our_range) +{ + hvb_page_range_tree_destroy(&our_range->removed_guest); + hvb_page_range_tree_destroy(&our_range->removed_both); + hvb_page_range_tree_init(&our_range->removed_guest); + hvb_page_range_tree_init(&our_range->removed_both); +} + +void hvb_our_range_mark_added(OurRange *our_range, uint64_t additional_size) +{ + assert(additional_size <= UINT64_MAX - our_range->added); + + our_range->added += additional_size; + + assert(our_range->added <= UINT64_MAX - our_range->unusable_tail); + assert(our_range->added + our_range->unusable_tail <= + our_range->range.count); +} + +/* OurRangeMemslots */ +static void our_range_memslots_init_slots(OurRangeMemslots *our_range, + MemoryRegion *backing_mr, + Object *memslot_owner) +{ + OurRangeMemslotsSlots *memslots = &our_range->slots; + unsigned int idx; + uint64_t memslot_offset; + + assert(memslots->count > 0); + memslots->slots = g_new0(MemoryRegion, memslots->count); + + /* Initialize our memslots, but don't map them yet. */ + assert(memslots->size_each > 0); + for (idx = 0, memslot_offset = 0; idx < memslots->count; + idx++, memslot_offset += memslots->size_each) { + uint64_t memslot_size; + g_autofree char *name = NULL; + + /* The size of the last memslot might be smaller. */ + if (idx == memslots->count - 1) { + uint64_t region_size; + + assert(our_range->mr); + region_size = memory_region_size(our_range->mr); + memslot_size = region_size - memslot_offset; + } else { + memslot_size = memslots->size_each; + } + + name = g_strdup_printf("memslot-%u", idx); + memory_region_init_alias(&memslots->slots[idx], memslot_owner, name, + backing_mr, memslot_offset, memslot_size); + /* + * We want to be able to atomically and efficiently activate/deactivate + * individual memslots without affecting adjacent memslots in memory + * notifiers. + */ + memory_region_set_unmergeable(&memslots->slots[idx], true); + } + + memslots->mapped_count = 0; +} + +OurRangeMemslots *hvb_our_range_memslots_new(uint64_t addr, + MemoryRegion *parent_mr, + MemoryRegion *backing_mr, + Object *memslot_owner, + unsigned int memslot_count, + uint64_t memslot_size) +{ + OurRangeMemslots *our_range; + + our_range = g_malloc(sizeof(*our_range)); + our_range_init(&our_range->range, + addr / HV_BALLOON_PAGE_SIZE, + memory_region_size(parent_mr) / HV_BALLOON_PAGE_SIZE); + our_range->slots.size_each = memslot_size; + our_range->slots.count = memslot_count; + our_range->mr = parent_mr; + our_range_memslots_init_slots(our_range, backing_mr, memslot_owner); + + return our_range; +} + +static void our_range_memslots_free_memslots(OurRangeMemslots *our_range) +{ + OurRangeMemslotsSlots *memslots = &our_range->slots; + unsigned int idx; + uint64_t offset; + + memory_region_transaction_begin(); + for (idx = 0, offset = 0; idx < memslots->mapped_count; + idx++, offset += memslots->size_each) { + trace_hv_balloon_unmap_slot(idx, memslots->count, offset); + assert(memory_region_is_mapped(&memslots->slots[idx])); + memory_region_del_subregion(our_range->mr, &memslots->slots[idx]); + } + memory_region_transaction_commit(); + + for (idx = 0; idx < memslots->count; idx++) { + object_unparent(OBJECT(&memslots->slots[idx])); + } + + g_clear_pointer(&our_range->slots.slots, g_free); +} + +void hvb_our_range_memslots_free(OurRangeMemslots *our_range) +{ + OurRangeMemslotsSlots *memslots = &our_range->slots; + MemoryRegion *hostmem_mr; + RAMBlock *rb; + + assert(our_range->slots.count > 0); + assert(our_range->slots.slots); + + hostmem_mr = memslots->slots[0].alias; + rb = hostmem_mr->ram_block; + ram_block_discard_range(rb, 0, qemu_ram_get_used_length(rb)); + + our_range_memslots_free_memslots(our_range); + our_range_destroy(&our_range->range); + g_free(our_range); +} + +void hvb_our_range_memslots_ensure_mapped_additional(OurRangeMemslots *our_range, + uint64_t additional_map_size) +{ + OurRangeMemslotsSlots *memslots = &our_range->slots; + uint64_t total_map_size; + unsigned int idx; + uint64_t offset; + + total_map_size = (our_range->range.added + additional_map_size) * + HV_BALLOON_PAGE_SIZE; + idx = memslots->mapped_count; + assert(memslots->size_each > 0); + offset = idx * memslots->size_each; + + /* + * Activate all memslots covered by the newly added region in a single + * transaction. + */ + memory_region_transaction_begin(); + for ( ; idx < memslots->count; + idx++, offset += memslots->size_each) { + /* + * If this memslot starts beyond or at the end of the range to map so + * does every next one. + */ + if (offset >= total_map_size) { + break; + } + + /* + * Instead of enabling/disabling memslot, we add/remove them. This + * should make address space updates faster, because we don't have to + * loop over many disabled subregions. + */ + trace_hv_balloon_map_slot(idx, memslots->count, offset); + assert(!memory_region_is_mapped(&memslots->slots[idx])); + memory_region_add_subregion(our_range->mr, offset, + &memslots->slots[idx]); + + memslots->mapped_count++; + } + memory_region_transaction_commit(); +} diff --git a/hw/hyperv/hv-balloon-our_range_memslots.h b/hw/hyperv/hv-balloon-our_range_memslots.h new file mode 100644 index 0000000..b6f592d --- /dev/null +++ b/hw/hyperv/hv-balloon-our_range_memslots.h @@ -0,0 +1,110 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2020-2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#ifndef HW_HYPERV_HV_BALLOON_OUR_RANGE_MEMSLOTS_H +#define HW_HYPERV_HV_BALLOON_OUR_RANGE_MEMSLOTS_H + +#include "qemu/osdep.h" + +#include "exec/memory.h" +#include "qom/object.h" +#include "hv-balloon-page_range_tree.h" + +/* OurRange */ +#define OUR_RANGE(ptr) ((OurRange *)(ptr)) + +/* "our range" means the memory range owned by this driver (for hot-adding) */ +typedef struct OurRange { + PageRange range; + + /* How many pages were hot-added to the guest */ + uint64_t added; + + /* Pages at the end not currently usable */ + uint64_t unusable_tail; + + /* Memory removed from the guest */ + PageRangeTree removed_guest, removed_both; +} OurRange; + +static inline uint64_t our_range_get_remaining_start(OurRange *our_range) +{ + return our_range->range.start + our_range->added; +} + +static inline uint64_t our_range_get_remaining_size(OurRange *our_range) +{ + return our_range->range.count - our_range->added - our_range->unusable_tail; +} + +void hvb_our_range_mark_added(OurRange *our_range, uint64_t additional_size); + +static inline void our_range_mark_remaining_unusable(OurRange *our_range) +{ + our_range->unusable_tail = our_range->range.count - our_range->added; +} + +static inline PageRangeTree our_range_get_removed_tree(OurRange *our_range, + bool both) +{ + if (both) { + return our_range->removed_both; + } else { + return our_range->removed_guest; + } +} + +static inline bool our_range_is_removed_tree_empty(OurRange *our_range, + bool both) +{ + if (both) { + return page_range_tree_is_empty(our_range->removed_both); + } else { + return page_range_tree_is_empty(our_range->removed_guest); + } +} + +void hvb_our_range_clear_removed_trees(OurRange *our_range); + +/* OurRangeMemslots */ +typedef struct OurRangeMemslotsSlots { + /* Nominal size of each memslot (the last one might be smaller) */ + uint64_t size_each; + + /* Slots array and its element count */ + MemoryRegion *slots; + unsigned int count; + + /* How many slots are currently mapped */ + unsigned int mapped_count; +} OurRangeMemslotsSlots; + +typedef struct OurRangeMemslots { + OurRange range; + + /* Memslots covering our range */ + OurRangeMemslotsSlots slots; + + MemoryRegion *mr; +} OurRangeMemslots; + +OurRangeMemslots *hvb_our_range_memslots_new(uint64_t addr, + MemoryRegion *parent_mr, + MemoryRegion *backing_mr, + Object *memslot_owner, + unsigned int memslot_count, + uint64_t memslot_size); +void hvb_our_range_memslots_free(OurRangeMemslots *our_range); + +G_DEFINE_AUTOPTR_CLEANUP_FUNC(OurRangeMemslots, hvb_our_range_memslots_free) + +void hvb_our_range_memslots_ensure_mapped_additional(OurRangeMemslots *our_range, + uint64_t additional_map_size); + +#endif diff --git a/hw/hyperv/hv-balloon.c b/hw/hyperv/hv-balloon.c index 5f3674b..5999f11 100644 --- a/hw/hyperv/hv-balloon.c +++ b/hw/hyperv/hv-balloon.c @@ -34,9 +34,12 @@ #include "sysemu/balloon.h" #include "sysemu/hostmem.h" #include "sysemu/reset.h" +#include "hv-balloon-our_range_memslots.h" #include "hv-balloon-page_range_tree.h" #include "trace.h" +#define HV_BALLOON_ADDR_PROP "addr" +#define HV_BALLOON_MEMDEV_PROP "memdev" #define HV_BALLOON_GUID "525074DC-8985-46e2-8057-A307DC18A502" /* @@ -52,6 +55,8 @@ #define HV_BALLOON_HA_CHUNK_SIZE (2 * GiB) #define HV_BALLOON_HA_CHUNK_PAGES (HV_BALLOON_HA_CHUNK_SIZE / HV_BALLOON_PAGE_SIZE) +#define HV_BALLOON_HA_MEMSLOT_SIZE_ALIGN (128 * MiB) + #define HV_BALLOON_HR_CHUNK_PAGES 585728 /* * ^ that's the maximum number of pages @@ -88,6 +93,10 @@ typedef enum State { S_UNBALLOON_POSTING, S_UNBALLOON_RB_WAIT, S_UNBALLOON_REPLY_WAIT, + S_HOT_ADD_SETUP, + S_HOT_ADD_RB_WAIT, + S_HOT_ADD_POSTING, + S_HOT_ADD_REPLY_WAIT, } State; typedef struct StateDesc { @@ -117,25 +126,43 @@ typedef struct HvBalloon { uint64_t target; bool target_changed; - /* Current (un)balloon */ + /* Current (un)balloon / hot-add operation parameters */ union { uint64_t balloon_diff; struct { uint64_t unballoon_diff; + uint64_t hot_add_diff; + }; + + struct { + PageRange hot_add_range; + uint64_t ha_current_count; }; }; + OurRangeMemslots *our_range; + + /* Count of memslots covering our memory */ + unsigned int memslot_count; + /* Nominal size of each memslot (the last one might be smaller) */ uint64_t memslot_size; + /* Non-ours removed memory */ PageRangeTree removed_guest, removed_both; + /* Grand totals of removed memory (both ours and non-ours) */ uint64_t removed_guest_ctr, removed_both_ctr; + + /* MEMORY_DEVICE props */ + uint64_t addr; + HostMemoryBackend *hostmem; + MemoryRegion *mr; } HvBalloon; OBJECT_DEFINE_TYPE_WITH_INTERFACES(HvBalloon, hv_balloon, HV_BALLOON, VMBUS_DEVICE, \ - { }) + { TYPE_MEMORY_DEVICE }, { }) #define HV_BALLOON_SET_STATE(hvb, news) \ do { \ @@ -155,6 +182,16 @@ typedef struct HvBalloonReq { VMBusChanReq vmreq; } HvBalloonReq; +/* total our memory includes parts currently removed from the guest */ +static uint64_t hv_balloon_total_our_ram(HvBalloon *balloon) +{ + if (!balloon->our_range) { + return 0; + } + + return balloon->our_range->range.added; +} + /* TODO: unify the code below with virtio-balloon and cache the value */ static int build_dimm_list(Object *obj, void *opaque) { @@ -193,10 +230,11 @@ static uint64_t hv_balloon_total_ram(HvBalloon *balloon) { ram_addr_t ram_size = get_current_ram_size(); uint64_t ram_size_pages = ram_size >> HV_BALLOON_PFN_SHIFT; + uint64_t our_ram_size_pages = hv_balloon_total_our_ram(balloon); assert(ram_size_pages > 0); - return ram_size_pages; + return SUM_SATURATE_U64(ram_size_pages, our_ram_size_pages); } /* @@ -275,14 +313,30 @@ static ssize_t hv_balloon_send_packet(VMBusChannel *chan, static bool hv_balloon_unballoon_get_source(HvBalloon *balloon, PageRangeTree *dtree, - uint64_t **dctr) + uint64_t **dctr, + bool *is_our_range) { + OurRange *our_range = OUR_RANGE(balloon->our_range); + + /* Try the boot memory first */ if (g_tree_nnodes(balloon->removed_guest.t) > 0) { *dtree = balloon->removed_guest; *dctr = &balloon->removed_guest_ctr; + *is_our_range = false; } else if (g_tree_nnodes(balloon->removed_both.t) > 0) { *dtree = balloon->removed_both; *dctr = &balloon->removed_both_ctr; + *is_our_range = false; + } else if (!our_range) { + return false; + } else if (!our_range_is_removed_tree_empty(our_range, false)) { + *dtree = our_range_get_removed_tree(our_range, false); + *dctr = &balloon->removed_guest_ctr; + *is_our_range = true; + } else if (!our_range_is_removed_tree_empty(our_range, true)) { + *dtree = our_range_get_removed_tree(our_range, true); + *dctr = &balloon->removed_both_ctr; + *is_our_range = true; } else { return false; } @@ -310,6 +364,7 @@ static void hv_balloon_unballoon_posting(HvBalloon *balloon, StateDesc *stdesc) VMBusChannel *chan = hv_balloon_get_channel(balloon); PageRangeTree dtree; uint64_t *dctr; + bool our_range; struct dm_unballoon_request *ur; size_t ur_size = sizeof(*ur) + sizeof(ur->range_array[0]); PageRange range; @@ -319,7 +374,7 @@ static void hv_balloon_unballoon_posting(HvBalloon *balloon, StateDesc *stdesc) assert(balloon->state == S_UNBALLOON_POSTING); assert(balloon->unballoon_diff > 0); - if (!hv_balloon_unballoon_get_source(balloon, &dtree, &dctr)) { + if (!hv_balloon_unballoon_get_source(balloon, &dtree, &dctr, &our_range)) { error_report("trying to unballoon but nothing seems to be ballooned"); /* * there is little we can do as we might have already @@ -328,6 +383,7 @@ static void hv_balloon_unballoon_posting(HvBalloon *balloon, StateDesc *stdesc) return; } + assert(balloon->our_range || !our_range); assert(dtree.t); assert(dctr); @@ -369,6 +425,166 @@ static void hv_balloon_unballoon_posting(HvBalloon *balloon, StateDesc *stdesc) } } +static bool hv_balloon_our_range_ensure(HvBalloon *balloon) +{ + uint64_t align; + MemoryRegion *hostmem_mr; + g_autoptr(OurRangeMemslots) our_range_memslots = NULL; + OurRange *our_range; + + if (balloon->our_range) { + return true; + } + + if (!balloon->hostmem) { + return false; + } + + align = (1 << balloon->caps.cap_bits.hot_add_alignment) * MiB; + assert(QEMU_IS_ALIGNED(balloon->addr, align)); + + hostmem_mr = host_memory_backend_get_memory(balloon->hostmem); + + our_range_memslots = hvb_our_range_memslots_new(balloon->addr, + balloon->mr, hostmem_mr, + OBJECT(balloon), + balloon->memslot_count, + balloon->memslot_size); + our_range = OUR_RANGE(our_range_memslots); + + if (hvb_page_range_tree_intree_any(balloon->removed_guest, + our_range->range.start, + our_range->range.count) || + hvb_page_range_tree_intree_any(balloon->removed_both, + our_range->range.start, + our_range->range.count)) { + error_report("some parts of the memory backend were already returned by the guest. this should not happen, please reboot the guest and try again"); + return false; + } + + trace_hv_balloon_our_range_add(our_range->range.count, + our_range->range.start); + + balloon->our_range = g_steal_pointer(&our_range_memslots); + return true; +} + +static void hv_balloon_hot_add_setup(HvBalloon *balloon, StateDesc *stdesc) +{ + /* need to make copy since it is in union with hot_add_range */ + uint64_t hot_add_diff = balloon->hot_add_diff; + PageRange *hot_add_range = &balloon->hot_add_range; + uint64_t align, our_range_remaining; + OurRange *our_range; + + assert(balloon->state == S_HOT_ADD_SETUP); + assert(hot_add_diff > 0); + + if (!hv_balloon_our_range_ensure(balloon)) { + goto ret_idle; + } + + our_range = OUR_RANGE(balloon->our_range); + + align = (1 << balloon->caps.cap_bits.hot_add_alignment) * + (MiB / HV_BALLOON_PAGE_SIZE); + + /* Absolute GPA in pages */ + hot_add_range->start = our_range_get_remaining_start(our_range); + assert(QEMU_IS_ALIGNED(hot_add_range->start, align)); + + our_range_remaining = our_range_get_remaining_size(our_range); + hot_add_range->count = MIN(our_range_remaining, hot_add_diff); + hot_add_range->count = QEMU_ALIGN_DOWN(hot_add_range->count, align); + if (hot_add_range->count == 0) { + goto ret_idle; + } + + hvb_our_range_memslots_ensure_mapped_additional(balloon->our_range, + hot_add_range->count); + + HV_BALLOON_STATE_DESC_SET(stdesc, S_HOT_ADD_RB_WAIT); + return; + +ret_idle: + HV_BALLOON_STATE_DESC_SET(stdesc, S_IDLE); +} + +static void hv_balloon_hot_add_rb_wait(HvBalloon *balloon, StateDesc *stdesc) +{ + VMBusChannel *chan = hv_balloon_get_channel(balloon); + struct dm_hot_add *ha; + size_t ha_size = sizeof(*ha) + sizeof(ha->range); + + assert(balloon->state == S_HOT_ADD_RB_WAIT); + + if (vmbus_channel_reserve(chan, 0, ha_size) < 0) { + return; + } + + HV_BALLOON_STATE_DESC_SET(stdesc, S_HOT_ADD_POSTING); +} + +static void hv_balloon_hot_add_posting(HvBalloon *balloon, StateDesc *stdesc) +{ + PageRange *hot_add_range = &balloon->hot_add_range; + uint64_t *current_count = &balloon->ha_current_count; + VMBusChannel *chan = hv_balloon_get_channel(balloon); + struct dm_hot_add *ha; + size_t ha_size = sizeof(*ha) + sizeof(ha->range); + union dm_mem_page_range *ha_region; + uint64_t align, chunk_max_size; + ssize_t ret; + + assert(balloon->state == S_HOT_ADD_POSTING); + assert(hot_add_range->count > 0); + + align = (1 << balloon->caps.cap_bits.hot_add_alignment) * + (MiB / HV_BALLOON_PAGE_SIZE); + if (align >= HV_BALLOON_HA_CHUNK_PAGES) { + /* + * If the required alignment is higher than the chunk size we let it + * override that size. + */ + chunk_max_size = align; + } else { + chunk_max_size = QEMU_ALIGN_DOWN(HV_BALLOON_HA_CHUNK_PAGES, align); + } + + /* + * hot_add_range->count starts aligned in hv_balloon_hot_add_setup(), + * then it is either reduced by subtracting aligned current_count or + * further hot-adds are prevented by marking the whole remaining our range + * as unusable in hv_balloon_handle_hot_add_response(). + */ + *current_count = MIN(hot_add_range->count, chunk_max_size); + + ha = alloca(ha_size); + ha_region = &(&ha->range)[1]; + memset(ha, 0, ha_size); + ha->hdr.type = DM_MEM_HOT_ADD_REQUEST; + ha->hdr.size = ha_size; + ha->hdr.trans_id = balloon->trans_id; + + ha->range.finfo.start_page = hot_add_range->start; + ha->range.finfo.page_cnt = *current_count; + ha_region->finfo.start_page = hot_add_range->start; + ha_region->finfo.page_cnt = ha->range.finfo.page_cnt; + + trace_hv_balloon_outgoing_hot_add(ha->hdr.trans_id, + *current_count, hot_add_range->start); + + ret = vmbus_channel_send(chan, VMBUS_PACKET_DATA_INBAND, + NULL, 0, ha, ha_size, false, + ha->hdr.trans_id); + if (ret <= 0) { + error_report("error %zd when posting hot add msg, expect problems", + ret); + } + + HV_BALLOON_STATE_DESC_SET(stdesc, S_HOT_ADD_REPLY_WAIT); +} + static void hv_balloon_balloon_rb_wait(HvBalloon *balloon, StateDesc *stdesc) { VMBusChannel *chan = hv_balloon_get_channel(balloon); @@ -428,14 +644,23 @@ static void hv_balloon_idle_state_process_target(HvBalloon *balloon, * the procedure is in progress */ if (balloon->target > ram_size_pages - total_removed) { + bool can_hot_add = balloon->caps.cap_bits.hot_add; uint64_t target_diff = balloon->target - (ram_size_pages - total_removed); balloon->unballoon_diff = MIN(target_diff, total_removed); + if (can_hot_add) { + balloon->hot_add_diff = target_diff - balloon->unballoon_diff; + } else { + balloon->hot_add_diff = 0; + } + if (balloon->unballoon_diff > 0) { assert(can_balloon); HV_BALLOON_STATE_DESC_SET(stdesc, S_UNBALLOON_RB_WAIT); + } else if (balloon->hot_add_diff > 0) { + HV_BALLOON_STATE_DESC_SET(stdesc, S_HOT_ADD_SETUP); } } else if (can_balloon && balloon->target < ram_size_pages - total_removed) { @@ -465,6 +690,9 @@ static const struct { [S_BALLOON_RB_WAIT].handler = hv_balloon_balloon_rb_wait, [S_UNBALLOON_POSTING].handler = hv_balloon_unballoon_posting, [S_UNBALLOON_RB_WAIT].handler = hv_balloon_unballoon_rb_wait, + [S_HOT_ADD_SETUP].handler = hv_balloon_hot_add_setup, + [S_HOT_ADD_RB_WAIT].handler = hv_balloon_hot_add_rb_wait, + [S_HOT_ADD_POSTING].handler = hv_balloon_hot_add_posting, }; static void hv_balloon_handle_state(HvBalloon *balloon, StateDesc *stdesc) @@ -507,13 +735,64 @@ static void hv_balloon_remove_response_handle_range(HvBalloon *balloon, bool both, uint64_t *removedctr) { + OurRange *our_range = OUR_RANGE(balloon->our_range); PageRangeTree globaltree = both ? balloon->removed_both : balloon->removed_guest; uint64_t *globalctr = both ? &balloon->removed_both_ctr : &balloon->removed_guest_ctr; + PageRange rangeeff; + + if (range->count == 0) { + return; + } trace_hv_balloon_remove_response(range->count, range->start, both); + if (our_range) { + /* Includes the not-yet-hot-added and unusable parts. */ + rangeeff = our_range->range; + } else { + rangeeff.start = rangeeff.count = 0; + } + + if (page_range_intersection_size(range, rangeeff.start, rangeeff.count) > 0) { + PageRangeTree ourtree = our_range_get_removed_tree(our_range, both); + PageRange rangehole, rangecommon; + uint64_t ourremoved = 0; + + /* process the hole before our range, if it exists */ + page_range_part_before(range, rangeeff.start, &rangehole); + hv_balloon_remove_response_insert_range(globaltree, &rangehole, + globalctr, removedctr, NULL); + if (rangehole.count > 0) { + trace_hv_balloon_remove_response_hole(rangehole.count, + rangehole.start, + range->count, range->start, + rangeeff.start, both); + } + + /* process our part */ + page_range_intersect(range, rangeeff.start, rangeeff.count, + &rangecommon); + hv_balloon_remove_response_insert_range(ourtree, &rangecommon, + globalctr, removedctr, + &ourremoved); + if (rangecommon.count > 0) { + trace_hv_balloon_remove_response_common(rangecommon.count, + rangecommon.start, + range->count, range->start, + rangeeff.count, + rangeeff.start, ourremoved, + both); + } + + /* calculate what's left after our range */ + rangecommon = *range; + page_range_part_after(&rangecommon, rangeeff.start, rangeeff.count, + range); + } + + /* process the remainder of the range that lies after our range */ if (range->count > 0) { hv_balloon_remove_response_insert_range(globaltree, range, globalctr, removedctr, NULL); @@ -847,6 +1126,72 @@ static void hv_balloon_handle_unballoon_response(HvBalloon *balloon, balloon->trans_id++; + if (balloon->hot_add_diff > 0) { + bool can_hot_add = balloon->caps.cap_bits.hot_add; + + assert(can_hot_add); + HV_BALLOON_STATE_DESC_SET(stdesc, S_HOT_ADD_SETUP); + } else { + HV_BALLOON_STATE_DESC_SET(stdesc, S_IDLE); + } +} + +static void hv_balloon_handle_hot_add_response(HvBalloon *balloon, + HvBalloonReq *req, + StateDesc *stdesc) +{ + PageRange *hot_add_range = &balloon->hot_add_range; + VMBusChanReq *vmreq = &req->vmreq; + struct dm_hot_add_response *msgHaR = vmreq->msg; + OurRange *our_range; + + if (balloon->state != S_HOT_ADD_REPLY_WAIT) { + warn_report("unexpected DM_HOT_ADD_RESPONSE in %d state", + balloon->state); + return; + } + + assert(balloon->our_range); + our_range = OUR_RANGE(balloon->our_range); + + if (!hv_balloon_handle_msg_size(req, sizeof(*msgHaR), + "DM_HOT_ADD_RESPONSE")) + return; + + trace_hv_balloon_incoming_hot_add(msgHaR->hdr.trans_id, msgHaR->result, + msgHaR->page_count); + + balloon->trans_id++; + + if (msgHaR->result) { + if (msgHaR->page_count > balloon->ha_current_count) { + warn_report("DM_HOT_ADD_RESPONSE page count higher than requested (%"PRIu32" vs %"PRIu64")", + msgHaR->page_count, balloon->ha_current_count); + msgHaR->page_count = balloon->ha_current_count; + } + + hvb_our_range_mark_added(our_range, msgHaR->page_count); + hot_add_range->start += msgHaR->page_count; + hot_add_range->count -= msgHaR->page_count; + } + + if (!msgHaR->result || msgHaR->page_count < balloon->ha_current_count) { + /* + * the current planned range was only partially hot-added, take note + * how much of it remains and don't attempt any further hot adds + */ + our_range_mark_remaining_unusable(our_range); + + goto ret_idle; + } + + /* any pages remaining to hot-add in our range? */ + if (hot_add_range->count > 0) { + HV_BALLOON_STATE_DESC_SET(stdesc, S_HOT_ADD_RB_WAIT); + return; + } + +ret_idle: HV_BALLOON_STATE_DESC_SET(stdesc, S_IDLE); } @@ -924,6 +1269,10 @@ static void hv_balloon_handle_packet(HvBalloon *balloon, HvBalloonReq *req, hv_balloon_handle_status_report(balloon, req); break; + case DM_MEM_HOT_ADD_RESPONSE: + hv_balloon_handle_hot_add_response(balloon, req, stdesc); + break; + case DM_UNBALLOON_RESPONSE: hv_balloon_handle_unballoon_response(balloon, req, stdesc); break; @@ -1072,6 +1421,61 @@ static void hv_balloon_post_init_timer(void *opaque) hv_balloon_event_loop(balloon); } +static void hv_balloon_system_reset_unrealize_common(HvBalloon *balloon) +{ + g_clear_pointer(&balloon->our_range, hvb_our_range_memslots_free); +} + +static void hv_balloon_system_reset(void *opaque) +{ + HvBalloon *balloon = HV_BALLOON(opaque); + + hv_balloon_system_reset_unrealize_common(balloon); +} + +static void hv_balloon_ensure_mr(HvBalloon *balloon) +{ + MemoryRegion *hostmem_mr; + + assert(balloon->hostmem); + + if (balloon->mr) { + return; + } + + hostmem_mr = host_memory_backend_get_memory(balloon->hostmem); + + balloon->mr = g_new0(MemoryRegion, 1); + memory_region_init(balloon->mr, OBJECT(balloon), TYPE_HV_BALLOON, + memory_region_size(hostmem_mr)); + + /* + * The VM can indicate an alignment up to 32 GiB. Memory device core can + * usually only handle/guarantee 1 GiB alignment. The user will have to + * specify a larger maxmem eventually. + * + * The memory device core will warn the user in case maxmem might have to be + * increased and will fail plugging the device if there is not sufficient + * space after alignment. + * + * TODO: we could do the alignment ourselves in a slightly bigger region. + * But this feels better, although the warning might be annoying. Maybe + * we can optimize that in the future (e.g., with such a device on the + * cmdline place/size the device memory region differently. + */ + balloon->mr->align = MAX(32 * GiB, memory_region_get_alignment(hostmem_mr)); +} + +static void hv_balloon_free_mr(HvBalloon *balloon) +{ + if (!balloon->mr) { + return; + } + + object_unparent(OBJECT(balloon->mr)); + g_clear_pointer(&balloon->mr, g_free); +} + static void hv_balloon_vmdev_realize(VMBusDevice *vdev, Error **errp) { ERRP_GUARD(); @@ -1088,13 +1492,52 @@ static void hv_balloon_vmdev_realize(VMBusDevice *vdev, Error **errp) return; } + if (balloon->hostmem) { + if (host_memory_backend_is_mapped(balloon->hostmem)) { + Object *obj = OBJECT(balloon->hostmem); + + error_setg(errp, "'%s' property specifies a busy memdev: %s", + HV_BALLOON_MEMDEV_PROP, + object_get_canonical_path_component(obj)); + goto out_balloon_handler; + } + + hv_balloon_ensure_mr(balloon); + + /* This is rather unlikely to happen, but let's still check for it. */ + if (!QEMU_IS_ALIGNED(memory_region_size(balloon->mr), + HV_BALLOON_PAGE_SIZE)) { + error_setg(errp, "'%s' property memdev size has to be a multiple of 0x%" PRIx64, + HV_BALLOON_MEMDEV_PROP, (uint64_t)HV_BALLOON_PAGE_SIZE); + goto out_balloon_handler; + } + + host_memory_backend_set_mapped(balloon->hostmem, true); + vmstate_register_ram(host_memory_backend_get_memory(balloon->hostmem), + DEVICE(balloon)); + } else if (balloon->addr) { + error_setg(errp, "'%s' property must not be set without a memdev", + HV_BALLOON_MEMDEV_PROP); + goto out_balloon_handler; + } + timer_init_ms(&balloon->post_init_timer, QEMU_CLOCK_VIRTUAL, hv_balloon_post_init_timer, balloon); + + qemu_register_reset(hv_balloon_system_reset, balloon); + + return; + +out_balloon_handler: + qemu_remove_balloon_handler(balloon); } /* * VMBus device reset has to be implemented in case the guest decides to * disconnect and reconnect to the VMBus without rebooting the whole system. + * + * However, the hot-added memory can't be removed here as Windows keeps on using + * it until the system is restarted, even after disconnecting from the VMBus. */ static void hv_balloon_vmdev_reset(VMBusDevice *vdev) { @@ -1104,6 +1547,10 @@ static void hv_balloon_vmdev_reset(VMBusDevice *vdev) return; } + if (balloon->our_range) { + hvb_our_range_clear_removed_trees(OUR_RANGE(balloon->our_range)); + } + hvb_page_range_tree_destroy(&balloon->removed_guest); hvb_page_range_tree_destroy(&balloon->removed_both); hvb_page_range_tree_init(&balloon->removed_guest); @@ -1117,14 +1564,106 @@ static void hv_balloon_vmdev_reset(VMBusDevice *vdev) hv_balloon_event_loop(balloon); } +/* + * Clean up things that were (possibly) allocated pre-realization, for example + * from memory_device_pre_plug(), so we don't leak them if the device don't + * actually get realized in the end. + */ +static void hv_balloon_unrealize_finalize_common(HvBalloon *balloon) +{ + hv_balloon_free_mr(balloon); + balloon->addr = 0; + + balloon->memslot_count = 0; +} + static void hv_balloon_vmdev_unrealize(VMBusDevice *vdev) { HvBalloon *balloon = HV_BALLOON(vdev); + qemu_unregister_reset(hv_balloon_system_reset, balloon); + + hv_balloon_system_reset_unrealize_common(balloon); + qemu_remove_balloon_handler(balloon); + if (balloon->hostmem) { + vmstate_unregister_ram(host_memory_backend_get_memory(balloon->hostmem), + DEVICE(balloon)); + host_memory_backend_set_mapped(balloon->hostmem, false); + } + hvb_page_range_tree_destroy(&balloon->removed_guest); hvb_page_range_tree_destroy(&balloon->removed_both); + + hv_balloon_unrealize_finalize_common(balloon); +} + +static uint64_t hv_balloon_md_get_addr(const MemoryDeviceState *md) +{ + return object_property_get_uint(OBJECT(md), HV_BALLOON_ADDR_PROP, + &error_abort); +} + +static void hv_balloon_md_set_addr(MemoryDeviceState *md, uint64_t addr, + Error **errp) +{ + object_property_set_uint(OBJECT(md), HV_BALLOON_ADDR_PROP, addr, errp); +} + +static MemoryRegion *hv_balloon_md_get_memory_region(MemoryDeviceState *md, + Error **errp) +{ + HvBalloon *balloon = HV_BALLOON(md); + + if (!balloon->hostmem) { + return NULL; + } + + hv_balloon_ensure_mr(balloon); + + return balloon->mr; +} + +static void hv_balloon_decide_memslots(MemoryDeviceState *md, + unsigned int limit) +{ + HvBalloon *balloon = HV_BALLOON(md); + MemoryRegion *hostmem_mr; + uint64_t region_size, memslot_size, memslots; + + /* We're called exactly once, before realizing the device. */ + assert(!balloon->memslot_count); + + /* We should not be called if we don't have a memory backend */ + assert(balloon->hostmem); + + hostmem_mr = host_memory_backend_get_memory(balloon->hostmem); + region_size = memory_region_size(hostmem_mr); + + assert(region_size > 0); + memslot_size = QEMU_ALIGN_UP(region_size / limit, + HV_BALLOON_HA_MEMSLOT_SIZE_ALIGN); + memslots = QEMU_ALIGN_UP(region_size, memslot_size) / memslot_size; + + if (memslots > 1) { + balloon->memslot_size = memslot_size; + } else { + balloon->memslot_size = region_size; + } + + assert(memslots <= UINT_MAX); + balloon->memslot_count = memslots; +} + +static unsigned int hv_balloon_get_memslots(MemoryDeviceState *md) +{ + const HvBalloon *balloon = HV_BALLOON(md); + + /* We're called after setting the suggested limit. */ + assert(balloon->memslot_count > 0); + + return balloon->memslot_count; } static void hv_balloon_init(Object *obj) @@ -1133,12 +1672,20 @@ static void hv_balloon_init(Object *obj) static void hv_balloon_finalize(Object *obj) { + HvBalloon *balloon = HV_BALLOON(obj); + + hv_balloon_unrealize_finalize_common(balloon); } static Property hv_balloon_properties[] = { DEFINE_PROP_BOOL("status-report", HvBalloon, status_report.enabled, false), + /* MEMORY_DEVICE props */ + DEFINE_PROP_LINK(HV_BALLOON_MEMDEV_PROP, HvBalloon, hostmem, + TYPE_MEMORY_BACKEND, HostMemoryBackend *), + DEFINE_PROP_UINT64(HV_BALLOON_ADDR_PROP, HvBalloon, addr, 0), + DEFINE_PROP_END_OF_LIST(), }; @@ -1146,6 +1693,7 @@ static void hv_balloon_class_init(ObjectClass *klass, void *data) { DeviceClass *dc = DEVICE_CLASS(klass); VMBusDeviceClass *vdc = VMBUS_DEVICE_CLASS(klass); + MemoryDeviceClass *mdc = MEMORY_DEVICE_CLASS(klass); device_class_set_props(dc, hv_balloon_properties); qemu_uuid_parse(HV_BALLOON_GUID, &vdc->classid); @@ -1157,4 +1705,12 @@ static void hv_balloon_class_init(ObjectClass *klass, void *data) vdc->open_channel = hv_balloon_vmdev_open_channel; vdc->close_channel = hv_balloon_vmdev_close_channel; vdc->chan_notify_cb = hv_balloon_vmdev_chan_notify; + + mdc->get_addr = hv_balloon_md_get_addr; + mdc->set_addr = hv_balloon_md_set_addr; + mdc->get_plugged_size = memory_device_get_region_size; + mdc->get_memory_region = hv_balloon_md_get_memory_region; + mdc->decide_memslots = hv_balloon_decide_memslots; + mdc->get_memslots = hv_balloon_get_memslots; + /* implement fill_device_info */ } diff --git a/hw/hyperv/meson.build b/hw/hyperv/meson.build index b5be1cb..852d4f4 100644 --- a/hw/hyperv/meson.build +++ b/hw/hyperv/meson.build @@ -2,4 +2,4 @@ specific_ss.add(when: 'CONFIG_HYPERV', if_true: files('hyperv.c')) specific_ss.add(when: 'CONFIG_HYPERV_TESTDEV', if_true: files('hyperv_testdev.c')) specific_ss.add(when: 'CONFIG_VMBUS', if_true: files('vmbus.c')) specific_ss.add(when: 'CONFIG_SYNDBG', if_true: files('syndbg.c')) -specific_ss.add(when: 'CONFIG_HV_BALLOON', if_true: files('hv-balloon.c', 'hv-balloon-page_range_tree.c')) +specific_ss.add(when: 'CONFIG_HV_BALLOON', if_true: files('hv-balloon.c', 'hv-balloon-page_range_tree.c', 'hv-balloon-our_range_memslots.c')) diff --git a/hw/hyperv/trace-events b/hw/hyperv/trace-events index 86c724b..7963c21 100644 --- a/hw/hyperv/trace-events +++ b/hw/hyperv/trace-events @@ -23,9 +23,14 @@ hv_balloon_incoming_version(uint16_t major, uint16_t minor) "incoming proto vers hv_balloon_incoming_caps(uint32_t caps) "incoming caps 0x%x" hv_balloon_outgoing_unballoon(uint32_t trans_id, uint64_t count, uint64_t start, uint64_t rempages) "posting unballoon %"PRIu32" for %"PRIu64" @ 0x%"PRIx64", remaining %"PRIu64 hv_balloon_incoming_unballoon(uint32_t trans_id) "incoming unballoon response %"PRIu32 +hv_balloon_outgoing_hot_add(uint32_t trans_id, uint64_t count, uint64_t start) "posting hot add %"PRIu32" for %"PRIu64" @ 0x%"PRIx64 +hv_balloon_incoming_hot_add(uint32_t trans_id, uint32_t result, uint32_t count) "incoming hot add response %"PRIu32", result %"PRIu32", count %"PRIu32 hv_balloon_outgoing_balloon(uint32_t trans_id, uint64_t count, uint64_t rempages) "posting balloon %"PRIu32" for %"PRIu64", remaining %"PRIu64 hv_balloon_incoming_balloon(uint32_t trans_id, uint32_t range_count, uint32_t more_pages) "incoming balloon response %"PRIu32", ranges %"PRIu32", more %"PRIu32 +hv_balloon_our_range_add(uint64_t count, uint64_t start) "adding our range %"PRIu64" @ 0x%"PRIx64 hv_balloon_remove_response(uint64_t count, uint64_t start, unsigned int both) "processing remove response range %"PRIu64" @ 0x%"PRIx64", both %u" hv_balloon_remove_response_hole(uint64_t counthole, uint64_t starthole, uint64_t countrange, uint64_t startrange, uint64_t starthpr, unsigned int both) "response range hole %"PRIu64" @ 0x%"PRIx64" from range %"PRIu64" @ 0x%"PRIx64", before our start 0x%"PRIx64", both %u" hv_balloon_remove_response_common(uint64_t countcommon, uint64_t startcommon, uint64_t countrange, uint64_t startrange, uint64_t counthpr, uint64_t starthpr, uint64_t removed, unsigned int both) "response common range %"PRIu64" @ 0x%"PRIx64" from range %"PRIu64" @ 0x%"PRIx64" with our %"PRIu64" @ 0x%"PRIx64", removed %"PRIu64", both %u" hv_balloon_remove_response_remainder(uint64_t count, uint64_t start, unsigned int both) "remove response remaining range %"PRIu64" @ 0x%"PRIx64", both %u" +hv_balloon_map_slot(unsigned int idx, unsigned int total_slots, uint64_t offset) "mapping memslot %u / %u @ 0x%"PRIx64 +hv_balloon_unmap_slot(unsigned int idx, unsigned int total_slots, uint64_t offset) "unmapping memslot %u / %u @ 0x%"PRIx64 -- cgit v1.1 From 16dff2f9bb877bd1e147b5c5d9966d5a1d336c8c Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Wed, 23 Aug 2023 23:31:35 +0200 Subject: qapi: Add query-memory-devices support to hv-balloon Used by the driver to report its provided memory state information. Co-developed-by: David Hildenbrand Reviewed-by: David Hildenbrand Acked-by: Markus Armbruster Signed-off-by: Maciej S. Szmigiero --- hw/core/machine-hmp-cmds.c | 15 +++++++++++++++ hw/hyperv/hv-balloon.c | 27 ++++++++++++++++++++++++++- qapi/machine.json | 39 +++++++++++++++++++++++++++++++++++++-- 3 files changed, 78 insertions(+), 3 deletions(-) diff --git a/hw/core/machine-hmp-cmds.c b/hw/core/machine-hmp-cmds.c index 9a4b59c..a6ff6a4 100644 --- a/hw/core/machine-hmp-cmds.c +++ b/hw/core/machine-hmp-cmds.c @@ -253,6 +253,7 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict) MemoryDeviceInfo *value; PCDIMMDeviceInfo *di; SgxEPCDeviceInfo *se; + HvBalloonDeviceInfo *hi; for (info = info_list; info; info = info->next) { value = info->value; @@ -310,6 +311,20 @@ void hmp_info_memory_devices(Monitor *mon, const QDict *qdict) monitor_printf(mon, " node: %" PRId64 "\n", se->node); monitor_printf(mon, " memdev: %s\n", se->memdev); break; + case MEMORY_DEVICE_INFO_KIND_HV_BALLOON: + hi = value->u.hv_balloon.data; + monitor_printf(mon, "Memory device [%s]: \"%s\"\n", + MemoryDeviceInfoKind_str(value->type), + hi->id ? hi->id : ""); + if (hi->has_memaddr) { + monitor_printf(mon, " memaddr: 0x%" PRIx64 "\n", + hi->memaddr); + } + monitor_printf(mon, " max-size: %" PRIu64 "\n", hi->max_size); + if (hi->memdev) { + monitor_printf(mon, " memdev: %s\n", hi->memdev); + } + break; default: g_assert_not_reached(); } diff --git a/hw/hyperv/hv-balloon.c b/hw/hyperv/hv-balloon.c index 5999f11..44a8d15 100644 --- a/hw/hyperv/hv-balloon.c +++ b/hw/hyperv/hv-balloon.c @@ -1625,6 +1625,31 @@ static MemoryRegion *hv_balloon_md_get_memory_region(MemoryDeviceState *md, return balloon->mr; } +static void hv_balloon_md_fill_device_info(const MemoryDeviceState *md, + MemoryDeviceInfo *info) +{ + HvBalloonDeviceInfo *hi = g_new0(HvBalloonDeviceInfo, 1); + const HvBalloon *balloon = HV_BALLOON(md); + DeviceState *dev = DEVICE(md); + + if (dev->id) { + hi->id = g_strdup(dev->id); + } + + if (balloon->hostmem) { + hi->memdev = object_get_canonical_path(OBJECT(balloon->hostmem)); + hi->memaddr = balloon->addr; + hi->has_memaddr = true; + hi->max_size = memory_region_size(balloon->mr); + /* TODO: expose current provided size or something else? */ + } else { + hi->max_size = 0; + } + + info->u.hv_balloon.data = hi; + info->type = MEMORY_DEVICE_INFO_KIND_HV_BALLOON; +} + static void hv_balloon_decide_memslots(MemoryDeviceState *md, unsigned int limit) { @@ -1712,5 +1737,5 @@ static void hv_balloon_class_init(ObjectClass *klass, void *data) mdc->get_memory_region = hv_balloon_md_get_memory_region; mdc->decide_memslots = hv_balloon_decide_memslots; mdc->get_memslots = hv_balloon_get_memslots; - /* implement fill_device_info */ + mdc->fill_device_info = hv_balloon_md_fill_device_info; } diff --git a/qapi/machine.json b/qapi/machine.json index 6c9d2f6..2985d04 100644 --- a/qapi/machine.json +++ b/qapi/machine.json @@ -1290,6 +1290,29 @@ } ## +# @HvBalloonDeviceInfo: +# +# hv-balloon provided memory state information +# +# @id: device's ID +# +# @memaddr: physical address in memory, where device is mapped +# +# @max-size: the maximum size of memory that the device can provide +# +# @memdev: memory backend linked with device +# +# Since: 8.2 +## +{ 'struct': 'HvBalloonDeviceInfo', + 'data': { '*id': 'str', + '*memaddr': 'size', + 'max-size': 'size', + '*memdev': 'str' + } +} + +## # @MemoryDeviceInfoKind: # # @nvdimm: since 2.12 @@ -1300,10 +1323,13 @@ # # @sgx-epc: since 6.2. # +# @hv-balloon: since 8.2. +# # Since: 2.1 ## { 'enum': 'MemoryDeviceInfoKind', - 'data': [ 'dimm', 'nvdimm', 'virtio-pmem', 'virtio-mem', 'sgx-epc' ] } + 'data': [ 'dimm', 'nvdimm', 'virtio-pmem', 'virtio-mem', 'sgx-epc', + 'hv-balloon' ] } ## # @PCDIMMDeviceInfoWrapper: @@ -1338,6 +1364,14 @@ 'data': { 'data': 'SgxEPCDeviceInfo' } } ## +# @HvBalloonDeviceInfoWrapper: +# +# Since: 8.2 +## +{ 'struct': 'HvBalloonDeviceInfoWrapper', + 'data': { 'data': 'HvBalloonDeviceInfo' } } + +## # @MemoryDeviceInfo: # # Union containing information about a memory device @@ -1351,7 +1385,8 @@ 'nvdimm': 'PCDIMMDeviceInfoWrapper', 'virtio-pmem': 'VirtioPMEMDeviceInfoWrapper', 'virtio-mem': 'VirtioMEMDeviceInfoWrapper', - 'sgx-epc': 'SgxEPCDeviceInfoWrapper' + 'sgx-epc': 'SgxEPCDeviceInfoWrapper', + 'hv-balloon': 'HvBalloonDeviceInfoWrapper' } } -- cgit v1.1 From 259ebed45a2cd2ae6a5820fbc2e51578d2ad4eb7 Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Wed, 23 Aug 2023 23:34:14 +0200 Subject: qapi: Add HV_BALLOON_STATUS_REPORT event and its QMP query command Used by the hv-balloon driver for (optional) guest memory status reports. Acked-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- hw/hyperv/hv-balloon-stub.c | 19 ++++++++++++++ hw/hyperv/hv-balloon.c | 30 +++++++++++++++++++++- hw/hyperv/meson.build | 2 +- monitor/monitor.c | 1 + qapi/machine.json | 62 +++++++++++++++++++++++++++++++++++++++++++++ tests/qtest/qmp-cmd-test.c | 1 + 6 files changed, 113 insertions(+), 2 deletions(-) create mode 100644 hw/hyperv/hv-balloon-stub.c diff --git a/hw/hyperv/hv-balloon-stub.c b/hw/hyperv/hv-balloon-stub.c new file mode 100644 index 0000000..a47412d --- /dev/null +++ b/hw/hyperv/hv-balloon-stub.c @@ -0,0 +1,19 @@ +/* + * QEMU Hyper-V Dynamic Memory Protocol driver + * + * Copyright (C) 2023 Oracle and/or its affiliates. + * + * This work is licensed under the terms of the GNU GPL, version 2 or later. + * See the COPYING file in the top-level directory. + */ + +#include "qemu/osdep.h" +#include "qapi/error.h" +#include "qapi/qapi-commands-machine.h" +#include "qapi/qapi-types-machine.h" + +HvBalloonInfo *qmp_query_hv_balloon_status_report(Error **errp) +{ + error_setg(errp, "hv-balloon device not enabled in this build"); + return NULL; +} diff --git a/hw/hyperv/hv-balloon.c b/hw/hyperv/hv-balloon.c index 44a8d15..66f297c 100644 --- a/hw/hyperv/hv-balloon.c +++ b/hw/hyperv/hv-balloon.c @@ -1102,7 +1102,35 @@ static void hv_balloon_handle_status_report(HvBalloon *balloon, balloon->status_report.available *= HV_BALLOON_PAGE_SIZE; balloon->status_report.received = true; - /* report event */ + qapi_event_send_hv_balloon_status_report(balloon->status_report.committed, + balloon->status_report.available); +} + +HvBalloonInfo *qmp_query_hv_balloon_status_report(Error **errp) +{ + HvBalloon *balloon; + HvBalloonInfo *info; + + balloon = HV_BALLOON(object_resolve_path_type("", TYPE_HV_BALLOON, NULL)); + if (!balloon) { + error_setg(errp, "no %s device present", TYPE_HV_BALLOON); + return NULL; + } + + if (!balloon->status_report.enabled) { + error_setg(errp, "guest memory status reporting not enabled"); + return NULL; + } + + if (!balloon->status_report.received) { + error_setg(errp, "no guest memory status report received yet"); + return NULL; + } + + info = g_malloc0(sizeof(*info)); + info->committed = balloon->status_report.committed; + info->available = balloon->status_report.available; + return info; } static void hv_balloon_handle_unballoon_response(HvBalloon *balloon, diff --git a/hw/hyperv/meson.build b/hw/hyperv/meson.build index 852d4f4..d3d2668 100644 --- a/hw/hyperv/meson.build +++ b/hw/hyperv/meson.build @@ -2,4 +2,4 @@ specific_ss.add(when: 'CONFIG_HYPERV', if_true: files('hyperv.c')) specific_ss.add(when: 'CONFIG_HYPERV_TESTDEV', if_true: files('hyperv_testdev.c')) specific_ss.add(when: 'CONFIG_VMBUS', if_true: files('vmbus.c')) specific_ss.add(when: 'CONFIG_SYNDBG', if_true: files('syndbg.c')) -specific_ss.add(when: 'CONFIG_HV_BALLOON', if_true: files('hv-balloon.c', 'hv-balloon-page_range_tree.c', 'hv-balloon-our_range_memslots.c')) +specific_ss.add(when: 'CONFIG_HV_BALLOON', if_true: files('hv-balloon.c', 'hv-balloon-page_range_tree.c', 'hv-balloon-our_range_memslots.c'), if_false: files('hv-balloon-stub.c')) diff --git a/monitor/monitor.c b/monitor/monitor.c index 941f878..01ede1b 100644 --- a/monitor/monitor.c +++ b/monitor/monitor.c @@ -315,6 +315,7 @@ static MonitorQAPIEventConf monitor_qapi_event_conf[QAPI_EVENT__MAX] = { [QAPI_EVENT_QUORUM_FAILURE] = { 1000 * SCALE_MS }, [QAPI_EVENT_VSERPORT_CHANGE] = { 1000 * SCALE_MS }, [QAPI_EVENT_MEMORY_DEVICE_SIZE_CHANGE] = { 1000 * SCALE_MS }, + [QAPI_EVENT_HV_BALLOON_STATUS_REPORT] = { 1000 * SCALE_MS }, }; /* diff --git a/qapi/machine.json b/qapi/machine.json index 2985d04..b6d634b 100644 --- a/qapi/machine.json +++ b/qapi/machine.json @@ -1138,6 +1138,68 @@ 'data': { 'actual': 'int' } } ## +# @HvBalloonInfo: +# +# hv-balloon guest-provided memory status information. +# +# @committed: the amount of memory in use inside the guest plus the +# amount of the memory unusable inside the guest (ballooned out, +# offline, etc.) +# +# @available: the amount of the memory inside the guest available for +# new allocations ("free") +# +# Since: 8.2 +## +{ 'struct': 'HvBalloonInfo', + 'data': { 'committed': 'size', 'available': 'size' } } + +## +# @query-hv-balloon-status-report: +# +# Returns the hv-balloon driver data contained in the last received "STATUS" +# message from the guest. +# +# Returns: +# - @HvBalloonInfo on success +# - If no hv-balloon device is present, guest memory status reporting +# is not enabled or no guest memory status report received yet, +# GenericError +# +# Since: 8.2 +# +# Example: +# +# -> { "execute": "query-hv-balloon-status-report" } +# <- { "return": { +# "committed": 816640000, +# "available": 3333054464 +# } +# } +## +{ 'command': 'query-hv-balloon-status-report', 'returns': 'HvBalloonInfo' } + +## +# @HV_BALLOON_STATUS_REPORT: +# +# Emitted when the hv-balloon driver receives a "STATUS" message from +# the guest. +# +# Note: this event is rate-limited. +# +# Since: 8.2 +# +# Example: +# +# <- { "event": "HV_BALLOON_STATUS_REPORT", +# "data": { "committed": 816640000, "available": 3333054464 }, +# "timestamp": { "seconds": 1600295492, "microseconds": 661044 } } +# +## +{ 'event': 'HV_BALLOON_STATUS_REPORT', + 'data': 'HvBalloonInfo' } + +## # @MemoryInfo: # # Actual memory information in bytes. diff --git a/tests/qtest/qmp-cmd-test.c b/tests/qtest/qmp-cmd-test.c index 73a670e..2c15f60 100644 --- a/tests/qtest/qmp-cmd-test.c +++ b/tests/qtest/qmp-cmd-test.c @@ -45,6 +45,7 @@ static int query_error_class(const char *cmd) { "query-acpi-ospm-status", ERROR_CLASS_GENERIC_ERROR }, { "query-balloon", ERROR_CLASS_DEVICE_NOT_ACTIVE }, { "query-hotpluggable-cpus", ERROR_CLASS_GENERIC_ERROR }, + { "query-hv-balloon-status-report", ERROR_CLASS_GENERIC_ERROR }, { "query-vm-generation-id", ERROR_CLASS_GENERIC_ERROR }, /* Only valid with a USB bus added */ { "x-query-usb", ERROR_CLASS_GENERIC_ERROR }, -- cgit v1.1 From 9a52aa40dc039c2e047d86886a2f66ce03e2b98c Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Wed, 23 Aug 2023 23:36:35 +0200 Subject: hw/i386/pc: Support hv-balloon Add the necessary plumbing for the hv-balloon driver to the PC machine. Co-developed-by: David Hildenbrand Reviewed-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- hw/i386/Kconfig | 1 + hw/i386/pc.c | 22 ++++++++++++++++++++++ 2 files changed, 23 insertions(+) diff --git a/hw/i386/Kconfig b/hw/i386/Kconfig index 94772c7..5585079 100644 --- a/hw/i386/Kconfig +++ b/hw/i386/Kconfig @@ -45,6 +45,7 @@ config PC select ACPI_VMGENID select VIRTIO_PMEM_SUPPORTED select VIRTIO_MEM_SUPPORTED + select HV_BALLOON_SUPPORTED config PC_PCI bool diff --git a/hw/i386/pc.c b/hw/i386/pc.c index 6031234..1aef21a 100644 --- a/hw/i386/pc.c +++ b/hw/i386/pc.c @@ -27,6 +27,7 @@ #include "hw/i386/pc.h" #include "hw/char/serial.h" #include "hw/char/parallel.h" +#include "hw/hyperv/hv-balloon.h" #include "hw/i386/fw_cfg.h" #include "hw/i386/vmport.h" #include "sysemu/cpus.h" @@ -57,6 +58,7 @@ #include "hw/i386/kvm/xen_evtchn.h" #include "hw/i386/kvm/xen_gnttab.h" #include "hw/i386/kvm/xen_xenstore.h" +#include "hw/mem/memory-device.h" #include "e820_memory_layout.h" #include "trace.h" #include CONFIG_DEVICES @@ -1422,6 +1424,21 @@ static void pc_memory_unplug(HotplugHandler *hotplug_dev, error_propagate(errp, local_err); } +static void pc_hv_balloon_pre_plug(HotplugHandler *hotplug_dev, + DeviceState *dev, Error **errp) +{ + /* The vmbus handler has no hotplug handler; we should never end up here. */ + g_assert(!dev->hotplugged); + memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL, + errp); +} + +static void pc_hv_balloon_plug(HotplugHandler *hotplug_dev, + DeviceState *dev, Error **errp) +{ + memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev)); +} + static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev, DeviceState *dev, Error **errp) { @@ -1452,6 +1469,8 @@ static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev, return; } pcms->iommu = dev; + } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) { + pc_hv_balloon_pre_plug(hotplug_dev, dev, errp); } } @@ -1464,6 +1483,8 @@ static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev, x86_cpu_plug(hotplug_dev, dev, errp); } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) { virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp); + } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) { + pc_hv_balloon_plug(hotplug_dev, dev, errp); } } @@ -1505,6 +1526,7 @@ static HotplugHandler *pc_get_hotplug_handler(MachineState *machine, object_dynamic_cast(OBJECT(dev), TYPE_CPU) || object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) || object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) || + object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON) || object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) { return HOTPLUG_HANDLER(machine); } -- cgit v1.1 From 00313b517d09c0b141fb32997791f911c28fd3ff Mon Sep 17 00:00:00 2001 From: "Maciej S. Szmigiero" Date: Sat, 21 Oct 2023 23:14:53 +0200 Subject: MAINTAINERS: Add an entry for Hyper-V Dynamic Memory Protocol Acked-by: David Hildenbrand Signed-off-by: Maciej S. Szmigiero --- MAINTAINERS | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/MAINTAINERS b/MAINTAINERS index 8e8a7d5..d4a480c 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -2656,6 +2656,14 @@ F: hw/usb/canokey.c F: hw/usb/canokey.h F: docs/system/devices/canokey.rst +Hyper-V Dynamic Memory Protocol +M: Maciej S. Szmigiero +S: Supported +F: hw/hyperv/hv-balloon*.c +F: hw/hyperv/hv-balloon*.h +F: include/hw/hyperv/dynmem-proto.h +F: include/hw/hyperv/hv-balloon.h + Subsystems ---------- Overall Audio backends -- cgit v1.1