From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:37219) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1V6LhJ-0001gf-Ip for qemu-devel@nongnu.org; Mon, 05 Aug 2013 10:23:53 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1V6LhB-00063r-3G for qemu-devel@nongnu.org; Mon, 05 Aug 2013 10:23:45 -0400 Received: from mail-wi0-x232.google.com ([2a00:1450:400c:c05::232]:45460) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1V6LhA-00063S-Rf for qemu-devel@nongnu.org; Mon, 05 Aug 2013 10:23:37 -0400 Received: by mail-wi0-f178.google.com with SMTP id j17so1588708wiw.5 for ; Mon, 05 Aug 2013 07:23:35 -0700 (PDT) Sender: Paolo Bonzini Message-ID: <51FFB562.3080007@redhat.com> Date: Mon, 05 Aug 2013 16:23:30 +0200 From: Paolo Bonzini MIME-Version: 1.0 References: <1375690789-25212-1-git-send-email-lersek@redhat.com> <1375690789-25212-4-git-send-email-lersek@redhat.com> In-Reply-To: <1375690789-25212-4-git-send-email-lersek@redhat.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH v2 for-qmp-1.6 3/4] dump: populate guest_phys_blocks List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Laszlo Ersek , qemu-devel@nongnu.org On 08/05/2013 10:19 AM, Laszlo Ersek wrote: > While the machine is paused, in guest_phys_blocks_append() we register a > one-shot MemoryListener, solely for the initial collection of the valid > guest-physical memory ranges that happens at client registration time. > > For each range that is reported to guest_phys_blocks_set_memory(), we > attempt to merge the range with adjacent (preceding, subsequent, or both) > ranges. We use two hash tables for this purpose, both indexing the same > ranges, just by different keys (guest-phys-start vs. guest-phys-end). > > Ranges can only be joined if they are contiguous in both guest-physical > address space, and contiguous in host virtual address space. > > The "maximal" ranges that remain in the end constitute the guest-physical > memory map that the dump will be based on. > > Related RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=981582 I would expect the ranges to be pretty large with MemoryListener, so there should be no need for merging (on IRC in fact you mentioned only 10 calls to region_add). So I would zap all the merging code. Apart from this detail, this is great work! Paolo > Signed-off-by: Laszlo Ersek > --- > include/sysemu/memory_mapping.h | 1 + > dump.c | 2 +- > memory_mapping.c | 135 +++++++++++++++++++++++++++++++++++++++ > 3 files changed, 137 insertions(+), 1 deletions(-) > > diff --git a/include/sysemu/memory_mapping.h b/include/sysemu/memory_mapping.h > index 53c2cd5..6723dc5 100644 > --- a/include/sysemu/memory_mapping.h > +++ b/include/sysemu/memory_mapping.h > @@ -66,6 +66,7 @@ void memory_mapping_list_init(MemoryMappingList *list); > > void guest_phys_blocks_free(GuestPhysBlockList *list); > void guest_phys_blocks_init(GuestPhysBlockList *list); > +void guest_phys_blocks_append(GuestPhysBlockList *list); > > void qemu_get_guest_memory_mapping(MemoryMappingList *list, Error **errp); > > diff --git a/dump.c b/dump.c > index 716fb1d..3fa33fc 100644 > --- a/dump.c > +++ b/dump.c > @@ -746,7 +746,7 @@ static int dump_init(DumpState *s, int fd, bool paging, bool has_filter, > s->length = length; > > guest_phys_blocks_init(&s->guest_phys_blocks); > - /* FILL LIST */ > + guest_phys_blocks_append(&s->guest_phys_blocks); > > s->start = get_start_block(s); > if (s->start == -1) { > diff --git a/memory_mapping.c b/memory_mapping.c > index c70505b..efaabf8 100644 > --- a/memory_mapping.c > +++ b/memory_mapping.c > @@ -11,9 +11,13 @@ > * > */ > > +#include > + > #include "cpu.h" > #include "exec/cpu-all.h" > #include "sysemu/memory_mapping.h" > +#include "exec/memory.h" > +#include "exec/address-spaces.h" > > static void memory_mapping_list_add_mapping_sorted(MemoryMappingList *list, > MemoryMapping *mapping) > @@ -182,6 +186,137 @@ void guest_phys_blocks_init(GuestPhysBlockList *list) > QTAILQ_INIT(&list->head); > } > > +typedef struct GuestPhysListener { > + GHashTable *by_target_start; > + GHashTable *by_target_end; > + MemoryListener listener; > +} GuestPhysListener; > + > +static void guest_phys_blocks_region_add(MemoryListener *listener, > + MemoryRegionSection *section) > +{ > + GuestPhysListener *g; > + uint64_t section_size; > + hwaddr target_start, target_end; > + uint8_t *host_addr; > + GuestPhysBlock *predecessor, *successor, *block; > + bool found; > + > + /* we only care about RAM */ > + if (!memory_region_is_ram(section->mr)) { > + return; > + } > + > + g = container_of(listener, GuestPhysListener, listener); > + section_size = int128_get64(section->size); > + target_start = section->offset_within_address_space; > + target_end = target_start + section_size; > + host_addr = memory_region_get_ram_ptr(section->mr) + > + section->offset_within_region; > + > + /* find continuity in guest physical address space */ > + predecessor = g_hash_table_lookup(g->by_target_end, &target_start); > + successor = g_hash_table_lookup(g->by_target_start, &target_end); > + > + /* we require continuity in host memory too */ > + if (predecessor != NULL) { > + hwaddr predecessor_size = predecessor->target_end - > + predecessor->target_start; > + if (predecessor->host_addr + predecessor_size != host_addr) { > + predecessor = NULL; > + } > + } > + if (successor != NULL > + && host_addr + section_size != successor->host_addr) { > + successor = NULL; > + } > + > + if (predecessor == NULL) { > + if (successor == NULL) { > + /* Isolated mapping, allocate it and add it to both tables. */ > + block = g_malloc0(sizeof *block); > + > + block->target_end = target_end; > + g_hash_table_insert(g->by_target_end, &block->target_end, block); > + } else { > + /* Mapping has successor only. Merge current into successor by > + * modifying successor's start. Successor's end doesn't change. > + */ > + block = successor; > + found = g_hash_table_steal(g->by_target_start, > + &block->target_start); > + g_assert(found); > + } > + block->target_start = target_start; > + block->host_addr = host_addr; > + g_hash_table_insert(g->by_target_start, &block->target_start, block); > + return; > + } > + > + if (successor != NULL) { > + /* Mapping has both predecessor and successor. Delete the successor > + * and expand the predecessor to cover all three. > + */ > + target_end = successor->target_end; > + > + found = g_hash_table_steal(g->by_target_end, &successor->target_end); > + g_assert(found); > + found = g_hash_table_steal(g->by_target_start, > + &successor->target_start); > + g_assert(found); > + > + g_free(successor); > + } > + /* otherwise, mapping has predecessor only */ > + > + /* Expand predecessor until @target_end. Predecessor's start doesn't > + * change. > + */ > + block = predecessor; > + found = g_hash_table_steal(g->by_target_end, &block->target_end); > + g_assert(found); > + > + block->target_end = target_end; > + g_hash_table_insert(g->by_target_end, &block->target_end, block); > +} > + > +static void guest_phys_block_link(gpointer key, gpointer value, > + gpointer user_data) > +{ > + GuestPhysBlock *block = value; > + GuestPhysBlockList *list = user_data; > + > + QTAILQ_INSERT_TAIL(&list->head, block, next); > + ++list->num; > +} > + > +void guest_phys_blocks_append(GuestPhysBlockList *list) > +{ > + GHashFunc hash_func; > + GEqualFunc equal_func; > + GuestPhysListener g = { 0 }; > + > + if (sizeof(hwaddr) == sizeof(uint64_t)) { > + hash_func = &g_int64_hash; > + equal_func = &g_int64_equal; > + } else { > + hash_func = &g_int_hash; > + equal_func = &g_int_equal; > + } > + > + g.by_target_start = g_hash_table_new(hash_func, equal_func); > + g.by_target_end = g_hash_table_new(hash_func, equal_func); > + > + g.listener.region_add = &guest_phys_blocks_region_add; > + memory_listener_register(&g.listener, &address_space_memory); > + memory_listener_unregister(&g.listener); > + > + g_hash_table_foreach(g.by_target_start, &guest_phys_block_link, list); > + > + g_hash_table_destroy(g.by_target_end); > + g_hash_table_destroy(g.by_target_start); > +} > + > static CPUState *find_paging_enabled_cpu(CPUState *start_cpu) > { > CPUState *cpu; >