From f77c5f38f49c71bc14cf1019ac92b0b95f572414 Mon Sep 17 00:00:00 2001 From: David Hildenbrand Date: Wed, 17 Jan 2024 14:55:53 +0100 Subject: [PATCH 1/3] hv-balloon: use get_min_alignment() to express 32 GiB alignment Let's implement the get_min_alignment() callback for memory devices, and copy for the device memory region the alignment of the host memory region. This mimics what virtio-mem does, and allows for re-introducing proper alignment checks for the memory region size (where we don't care about additional device requirements) in memory device core. Message-ID: <20240117135554.787344-2-david@redhat.com> Reviewed-by: Maciej S. Szmigiero Signed-off-by: David Hildenbrand --- hw/hyperv/hv-balloon.c | 37 +++++++++++++++++++++---------------- 1 file changed, 21 insertions(+), 16 deletions(-) diff --git a/hw/hyperv/hv-balloon.c b/hw/hyperv/hv-balloon.c index 0238365712..ade283335a 100644 --- a/hw/hyperv/hv-balloon.c +++ b/hw/hyperv/hv-balloon.c @@ -1477,22 +1477,7 @@ static void hv_balloon_ensure_mr(HvBalloon *balloon) 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)); + balloon->mr->align = memory_region_get_alignment(hostmem_mr); } static void hv_balloon_free_mr(HvBalloon *balloon) @@ -1654,6 +1639,25 @@ static MemoryRegion *hv_balloon_md_get_memory_region(MemoryDeviceState *md, return balloon->mr; } +static uint64_t hv_balloon_md_get_min_alignment(const MemoryDeviceState *md) +{ + /* + * 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. + */ + return 32 * GiB; +} + static void hv_balloon_md_fill_device_info(const MemoryDeviceState *md, MemoryDeviceInfo *info) { @@ -1766,5 +1770,6 @@ 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; + mdc->get_min_alignment = hv_balloon_md_get_min_alignment; mdc->fill_device_info = hv_balloon_md_fill_device_info; } From 540a1abbf0b243e4cfb4333c5d30a041f7080ba4 Mon Sep 17 00:00:00 2001 From: David Hildenbrand Date: Wed, 17 Jan 2024 14:55:54 +0100 Subject: [PATCH 2/3] memory-device: reintroduce memory region size check We used to check that the memory region size is multiples of the overall requested address alignment for the device memory address. We removed that check, because there are cases (i.e., hv-balloon) where devices unconditionally request an address alignment that has a very large alignment (i.e., 32 GiB), but the actual memory device size might not be multiples of that alignment. However, this change: (a) allows for some practically impossible DIMM sizes, like "1GB+1 byte". (b) allows for DIMMs that partially cover hugetlb pages, previously reported in [1]. Both scenarios don't make any sense: we might even waste memory. So let's reintroduce that check, but only check that the memory region size is multiples of the memory region alignment (i.e., page size, huge page size), but not any additional memory device requirements communicated using md->get_min_alignment(). The following examples now fail again as expected: (a) 1M with 2M THP qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \ -object memory-backend-ram,id=mem1,size=1M \ -device pc-dimm,id=dimm1,memdev=mem1 -> backend memory size must be multiple of 0x200000 (b) 1G+1byte qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \ -object memory-backend-ram,id=mem1,size=1073741825B \ -device pc-dimm,id=dimm1,memdev=mem1 -> backend memory size must be multiple of 0x200000 (c) Unliagned hugetlb size (2M) qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \ -object memory-backend-file,id=mem1,mem-path=/dev/hugepages/tmp,size=511M \ -device pc-dimm,id=dimm1,memdev=mem1 backend memory size must be multiple of 0x200000 (d) Unliagned hugetlb size (1G) qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \ -object memory-backend-file,id=mem1,mem-path=/dev/hugepages1G/tmp,size=2047M \ -device pc-dimm,id=dimm1,memdev=mem1 -> backend memory size must be multiple of 0x40000000 Note that this fix depends on a hv-balloon change to communicate its additional alignment requirements using get_min_alignment() instead of through the memory region. [1] https://lkml.kernel.org/r/f77d641d500324525ac036fe1827b3070de75fc1.1701088320.git.mprivozn@redhat.com Message-ID: <20240117135554.787344-3-david@redhat.com> Reported-by: Zhenyu Zhang Reported-by: Michal Privoznik Fixes: eb1b7c4bd413 ("memory-device: Drop size alignment check") Tested-by: Zhenyu Zhang Tested-by: Mario Casquero Reviewed-by: Maciej S. Szmigiero Signed-off-by: David Hildenbrand --- hw/mem/memory-device.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) diff --git a/hw/mem/memory-device.c b/hw/mem/memory-device.c index a1b1af26bc..e098585cda 100644 --- a/hw/mem/memory-device.c +++ b/hw/mem/memory-device.c @@ -374,6 +374,20 @@ void memory_device_pre_plug(MemoryDeviceState *md, MachineState *ms, goto out; } + /* + * We always want the memory region size to be multiples of the memory + * region alignment: for example, DIMMs with 1G+1byte size don't make + * any sense. Note that we don't check that the size is multiples + * of any additional alignment requirements the memory device might + * have when it comes to the address in physical address space. + */ + if (!QEMU_IS_ALIGNED(memory_region_size(mr), + memory_region_get_alignment(mr))) { + error_setg(errp, "backend memory size must be multiple of 0x%" + PRIx64, memory_region_get_alignment(mr)); + return; + } + if (legacy_align) { align = *legacy_align; } else { From 04accf43df83aa10f06f7dbda3ecf0db97f0c5a6 Mon Sep 17 00:00:00 2001 From: Mark Kanda Date: Wed, 31 Jan 2024 10:53:27 -0600 Subject: [PATCH 3/3] oslib-posix: initialize backend memory objects in parallel QEMU initializes preallocated backend memory as the objects are parsed from the command line. This is not optimal in some cases (e.g. memory spanning multiple NUMA nodes) because the memory objects are initialized in series. Allow the initialization to occur in parallel (asynchronously). In order to ensure optimal thread placement, asynchronous initialization requires prealloc context threads to be in use. Signed-off-by: Mark Kanda Message-ID: <20240131165327.3154970-2-mark.kanda@oracle.com> Tested-by: Mario Casquero Signed-off-by: David Hildenbrand --- backends/hostmem.c | 7 ++- hw/virtio/virtio-mem.c | 4 +- include/hw/qdev-core.h | 5 ++ include/qemu/osdep.h | 18 +++++- system/vl.c | 9 +++ util/oslib-posix.c | 131 +++++++++++++++++++++++++++++++---------- util/oslib-win32.c | 8 ++- 7 files changed, 145 insertions(+), 37 deletions(-) diff --git a/backends/hostmem.c b/backends/hostmem.c index 987f6f591e..81a72ce40b 100644 --- a/backends/hostmem.c +++ b/backends/hostmem.c @@ -20,6 +20,7 @@ #include "qom/object_interfaces.h" #include "qemu/mmap-alloc.h" #include "qemu/madvise.h" +#include "hw/qdev-core.h" #ifdef CONFIG_NUMA #include @@ -237,7 +238,7 @@ static void host_memory_backend_set_prealloc(Object *obj, bool value, uint64_t sz = memory_region_size(&backend->mr); if (!qemu_prealloc_mem(fd, ptr, sz, backend->prealloc_threads, - backend->prealloc_context, errp)) { + backend->prealloc_context, false, errp)) { return; } backend->prealloc = true; @@ -323,6 +324,7 @@ host_memory_backend_memory_complete(UserCreatable *uc, Error **errp) HostMemoryBackendClass *bc = MEMORY_BACKEND_GET_CLASS(uc); void *ptr; uint64_t sz; + bool async = !phase_check(PHASE_LATE_BACKENDS_CREATED); if (!bc->alloc) { return; @@ -402,7 +404,8 @@ host_memory_backend_memory_complete(UserCreatable *uc, Error **errp) if (backend->prealloc && !qemu_prealloc_mem(memory_region_get_fd(&backend->mr), ptr, sz, backend->prealloc_threads, - backend->prealloc_context, errp)) { + backend->prealloc_context, + async, errp)) { return; } } diff --git a/hw/virtio/virtio-mem.c b/hw/virtio/virtio-mem.c index 99ab989852..ffd119ebac 100644 --- a/hw/virtio/virtio-mem.c +++ b/hw/virtio/virtio-mem.c @@ -605,7 +605,7 @@ static int virtio_mem_set_block_state(VirtIOMEM *vmem, uint64_t start_gpa, int fd = memory_region_get_fd(&vmem->memdev->mr); Error *local_err = NULL; - if (!qemu_prealloc_mem(fd, area, size, 1, NULL, &local_err)) { + if (!qemu_prealloc_mem(fd, area, size, 1, NULL, false, &local_err)) { static bool warned; /* @@ -1248,7 +1248,7 @@ static int virtio_mem_prealloc_range_cb(VirtIOMEM *vmem, void *arg, int fd = memory_region_get_fd(&vmem->memdev->mr); Error *local_err = NULL; - if (!qemu_prealloc_mem(fd, area, size, 1, NULL, &local_err)) { + if (!qemu_prealloc_mem(fd, area, size, 1, NULL, false, &local_err)) { error_report_err(local_err); return -ENOMEM; } diff --git a/include/hw/qdev-core.h b/include/hw/qdev-core.h index d47536eadb..9228e96c87 100644 --- a/include/hw/qdev-core.h +++ b/include/hw/qdev-core.h @@ -1083,6 +1083,11 @@ typedef enum MachineInitPhase { */ PHASE_ACCEL_CREATED, + /* + * Late backend objects have been created and initialized. + */ + PHASE_LATE_BACKENDS_CREATED, + /* * machine_class->init has been called, thus creating any embedded * devices and validating machine properties. Devices created at diff --git a/include/qemu/osdep.h b/include/qemu/osdep.h index c9692cc314..7d359dabc4 100644 --- a/include/qemu/osdep.h +++ b/include/qemu/osdep.h @@ -680,6 +680,8 @@ typedef struct ThreadContext ThreadContext; * @area: start address of the are to preallocate * @sz: the size of the area to preallocate * @max_threads: maximum number of threads to use + * @tc: prealloc context threads pointer, NULL if not in use + * @async: request asynchronous preallocation, requires @tc * @errp: returns an error if this function fails * * Preallocate memory (populate/prefault page tables writable) for the virtual @@ -687,10 +689,24 @@ typedef struct ThreadContext ThreadContext; * each page in the area was faulted in writable at least once, for example, * after allocating file blocks for mapped files. * + * When setting @async, allocation might be performed asynchronously. + * qemu_finish_async_prealloc_mem() must be called to finish any asynchronous + * preallocation. + * * Return: true on success, else false setting @errp with error. */ bool qemu_prealloc_mem(int fd, char *area, size_t sz, int max_threads, - ThreadContext *tc, Error **errp); + ThreadContext *tc, bool async, Error **errp); + +/** + * qemu_finish_async_prealloc_mem: + * @errp: returns an error if this function fails + * + * Finish all outstanding asynchronous memory preallocation. + * + * Return: true on success, else false setting @errp with error. + */ +bool qemu_finish_async_prealloc_mem(Error **errp); /** * qemu_get_pid_name: diff --git a/system/vl.c b/system/vl.c index bb959cbc44..2a0bd08ff1 100644 --- a/system/vl.c +++ b/system/vl.c @@ -2013,6 +2013,14 @@ static void qemu_create_late_backends(void) object_option_foreach_add(object_create_late); + /* + * Wait for any outstanding memory prealloc from created memory + * backends to complete. + */ + if (!qemu_finish_async_prealloc_mem(&error_fatal)) { + exit(1); + } + if (tpm_init() < 0) { exit(1); } @@ -3699,6 +3707,7 @@ void qemu_init(int argc, char **argv) * over memory-backend-file objects). */ qemu_create_late_backends(); + phase_advance(PHASE_LATE_BACKENDS_CREATED); /* * Note: creates a QOM object, must run only after global and diff --git a/util/oslib-posix.c b/util/oslib-posix.c index 7c297003b9..3c379f96c2 100644 --- a/util/oslib-posix.c +++ b/util/oslib-posix.c @@ -42,6 +42,7 @@ #include "qemu/cutils.h" #include "qemu/units.h" #include "qemu/thread-context.h" +#include "qemu/main-loop.h" #ifdef CONFIG_LINUX #include @@ -63,11 +64,15 @@ struct MemsetThread; +static QLIST_HEAD(, MemsetContext) memset_contexts = + QLIST_HEAD_INITIALIZER(memset_contexts); + typedef struct MemsetContext { bool all_threads_created; bool any_thread_failed; struct MemsetThread *threads; int num_threads; + QLIST_ENTRY(MemsetContext) next; } MemsetContext; struct MemsetThread { @@ -412,19 +417,44 @@ static inline int get_memset_num_threads(size_t hpagesize, size_t numpages, return ret; } +static int wait_and_free_mem_prealloc_context(MemsetContext *context) +{ + int i, ret = 0, tmp; + + for (i = 0; i < context->num_threads; i++) { + tmp = (uintptr_t)qemu_thread_join(&context->threads[i].pgthread); + + if (tmp) { + ret = tmp; + } + } + g_free(context->threads); + g_free(context); + return ret; +} + static int touch_all_pages(char *area, size_t hpagesize, size_t numpages, - int max_threads, ThreadContext *tc, + int max_threads, ThreadContext *tc, bool async, bool use_madv_populate_write) { static gsize initialized = 0; - MemsetContext context = { - .num_threads = get_memset_num_threads(hpagesize, numpages, max_threads), - }; + MemsetContext *context = g_malloc0(sizeof(MemsetContext)); size_t numpages_per_thread, leftover; void *(*touch_fn)(void *); - int ret = 0, i = 0; + int ret, i = 0; char *addr = area; + /* + * Asynchronous preallocation is only allowed when using MADV_POPULATE_WRITE + * and prealloc context for thread placement. + */ + if (!use_madv_populate_write || !tc) { + async = false; + } + + context->num_threads = + get_memset_num_threads(hpagesize, numpages, max_threads); + if (g_once_init_enter(&initialized)) { qemu_mutex_init(&page_mutex); qemu_cond_init(&page_cond); @@ -432,8 +462,11 @@ static int touch_all_pages(char *area, size_t hpagesize, size_t numpages, } if (use_madv_populate_write) { - /* Avoid creating a single thread for MADV_POPULATE_WRITE */ - if (context.num_threads == 1) { + /* + * Avoid creating a single thread for MADV_POPULATE_WRITE when + * preallocating synchronously. + */ + if (context->num_threads == 1 && !async) { if (qemu_madvise(area, hpagesize * numpages, QEMU_MADV_POPULATE_WRITE)) { return -errno; @@ -445,50 +478,86 @@ static int touch_all_pages(char *area, size_t hpagesize, size_t numpages, touch_fn = do_touch_pages; } - context.threads = g_new0(MemsetThread, context.num_threads); - numpages_per_thread = numpages / context.num_threads; - leftover = numpages % context.num_threads; - for (i = 0; i < context.num_threads; i++) { - context.threads[i].addr = addr; - context.threads[i].numpages = numpages_per_thread + (i < leftover); - context.threads[i].hpagesize = hpagesize; - context.threads[i].context = &context; + context->threads = g_new0(MemsetThread, context->num_threads); + numpages_per_thread = numpages / context->num_threads; + leftover = numpages % context->num_threads; + for (i = 0; i < context->num_threads; i++) { + context->threads[i].addr = addr; + context->threads[i].numpages = numpages_per_thread + (i < leftover); + context->threads[i].hpagesize = hpagesize; + context->threads[i].context = context; if (tc) { - thread_context_create_thread(tc, &context.threads[i].pgthread, + thread_context_create_thread(tc, &context->threads[i].pgthread, "touch_pages", - touch_fn, &context.threads[i], + touch_fn, &context->threads[i], QEMU_THREAD_JOINABLE); } else { - qemu_thread_create(&context.threads[i].pgthread, "touch_pages", - touch_fn, &context.threads[i], + qemu_thread_create(&context->threads[i].pgthread, "touch_pages", + touch_fn, &context->threads[i], QEMU_THREAD_JOINABLE); } - addr += context.threads[i].numpages * hpagesize; + addr += context->threads[i].numpages * hpagesize; + } + + if (async) { + /* + * async requests currently require the BQL. Add it to the list and kick + * preallocation off during qemu_finish_async_prealloc_mem(). + */ + assert(bql_locked()); + QLIST_INSERT_HEAD(&memset_contexts, context, next); + return 0; } if (!use_madv_populate_write) { - sigbus_memset_context = &context; + sigbus_memset_context = context; } qemu_mutex_lock(&page_mutex); - context.all_threads_created = true; + context->all_threads_created = true; qemu_cond_broadcast(&page_cond); qemu_mutex_unlock(&page_mutex); - for (i = 0; i < context.num_threads; i++) { - int tmp = (uintptr_t)qemu_thread_join(&context.threads[i].pgthread); + ret = wait_and_free_mem_prealloc_context(context); + if (!use_madv_populate_write) { + sigbus_memset_context = NULL; + } + return ret; +} + +bool qemu_finish_async_prealloc_mem(Error **errp) +{ + int ret = 0, tmp; + MemsetContext *context, *next_context; + + /* Waiting for preallocation requires the BQL. */ + assert(bql_locked()); + if (QLIST_EMPTY(&memset_contexts)) { + return true; + } + + qemu_mutex_lock(&page_mutex); + QLIST_FOREACH(context, &memset_contexts, next) { + context->all_threads_created = true; + } + qemu_cond_broadcast(&page_cond); + qemu_mutex_unlock(&page_mutex); + + QLIST_FOREACH_SAFE(context, &memset_contexts, next, next_context) { + QLIST_REMOVE(context, next); + tmp = wait_and_free_mem_prealloc_context(context); if (tmp) { ret = tmp; } } - if (!use_madv_populate_write) { - sigbus_memset_context = NULL; + if (ret) { + error_setg_errno(errp, -ret, + "qemu_prealloc_mem: preallocating memory failed"); + return false; } - g_free(context.threads); - - return ret; + return true; } static bool madv_populate_write_possible(char *area, size_t pagesize) @@ -498,7 +567,7 @@ static bool madv_populate_write_possible(char *area, size_t pagesize) } bool qemu_prealloc_mem(int fd, char *area, size_t sz, int max_threads, - ThreadContext *tc, Error **errp) + ThreadContext *tc, bool async, Error **errp) { static gsize initialized; int ret; @@ -540,7 +609,7 @@ bool qemu_prealloc_mem(int fd, char *area, size_t sz, int max_threads, } /* touch pages simultaneously */ - ret = touch_all_pages(area, hpagesize, numpages, max_threads, tc, + ret = touch_all_pages(area, hpagesize, numpages, max_threads, tc, async, use_madv_populate_write); if (ret) { error_setg_errno(errp, -ret, diff --git a/util/oslib-win32.c b/util/oslib-win32.c index c4a5f05a49..b623830d62 100644 --- a/util/oslib-win32.c +++ b/util/oslib-win32.c @@ -265,7 +265,7 @@ int getpagesize(void) } bool qemu_prealloc_mem(int fd, char *area, size_t sz, int max_threads, - ThreadContext *tc, Error **errp) + ThreadContext *tc, bool async, Error **errp) { int i; size_t pagesize = qemu_real_host_page_size(); @@ -278,6 +278,12 @@ bool qemu_prealloc_mem(int fd, char *area, size_t sz, int max_threads, return true; } +bool qemu_finish_async_prealloc_mem(Error **errp) +{ + /* async prealloc not supported, there is nothing to finish */ + return true; +} + char *qemu_get_pid_name(pid_t pid) { /* XXX Implement me */