From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:33322) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1cX4sk-0003Ii-SA for qemu-devel@nongnu.org; Fri, 27 Jan 2017 06:39:56 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1cX4sj-0006ag-MK for qemu-devel@nongnu.org; Fri, 27 Jan 2017 06:39:54 -0500 Received: from mx1.redhat.com ([209.132.183.28]:45432) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1cX4sj-0006aQ-DD for qemu-devel@nongnu.org; Fri, 27 Jan 2017 06:39:53 -0500 Date: Fri, 27 Jan 2017 11:39:48 +0000 From: "Dr. David Alan Gilbert" Message-ID: <20170127113947.GC3323@work-vm> References: <20170116155113.21034-1-pbutsykin@virtuozzo.com> <20170116155113.21034-3-pbutsykin@virtuozzo.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170116155113.21034-3-pbutsykin@virtuozzo.com> Subject: Re: [Qemu-devel] [PATCH 2/2] migration: discard non-dirty ram pages after the start of postcopy List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Pavel Butsykin Cc: qemu-devel@nongnu.org, quintela@redhat.com, armbru@redhat.com, amit.shah@redhat.com, den@openvz.org * Pavel Butsykin (pbutsykin@virtuozzo.com) wrote: > After the start of postcopy migration there are some non-dirty pages which have > already been migrated. These pages are no longer needed on the source vm so that > we can free them and it doen't hurt to complete the migration. > > Signed-off-by: Pavel Butsykin > --- > include/migration/migration.h | 1 + > migration/migration.c | 2 ++ > migration/ram.c | 25 +++++++++++++++++++++++++ > 3 files changed, 28 insertions(+) > > diff --git a/include/migration/migration.h b/include/migration/migration.h > index d7bd404365..0d9b81545c 100644 > --- a/include/migration/migration.h > +++ b/include/migration/migration.h > @@ -279,6 +279,7 @@ int ram_postcopy_send_discard_bitmap(MigrationState *ms); > int ram_discard_range(MigrationIncomingState *mis, const char *block_name, > uint64_t start, size_t length); > int ram_postcopy_incoming_init(MigrationIncomingState *mis); > +void ram_postcopy_migrated_memory_discard(MigrationState *ms); > > /** > * @migrate_add_blocker - prevent migration from proceeding > diff --git a/migration/migration.c b/migration/migration.c > index 391db6f28b..20490ed020 100644 > --- a/migration/migration.c > +++ b/migration/migration.c > @@ -1662,6 +1662,8 @@ static int postcopy_start(MigrationState *ms, bool *old_vm_running) > */ > qemu_savevm_send_ping(ms->to_dst_file, 4); > > + ram_postcopy_migrated_memory_discard(ms); > + Did you intend this to be selected based on your capability flag? > ret = qemu_file_get_error(ms->to_dst_file); > if (ret) { > error_report("postcopy_start: Migration stream errored"); > diff --git a/migration/ram.c b/migration/ram.c > index b0322a0b5c..8a6b614b0d 100644 > --- a/migration/ram.c > +++ b/migration/ram.c > @@ -1546,6 +1546,31 @@ void ram_debug_dump_bitmap(unsigned long *todump, bool expected) > > /* **** functions for postcopy ***** */ > > +void ram_postcopy_migrated_memory_discard(MigrationState *ms) > +{ > + struct RAMBlock *block; > + unsigned long *bitmap = atomic_rcu_read(&migration_bitmap_rcu)->bmap; > + > + QLIST_FOREACH_RCU(block, &ram_list.blocks, next) { > + unsigned long first = block->offset >> TARGET_PAGE_BITS; > + unsigned long range = first + (block->used_length >> TARGET_PAGE_BITS); > + unsigned long run_start = find_next_zero_bit(bitmap, range, first); > + > + while (run_start < range) { > + unsigned long run_end = find_next_bit(bitmap, range, run_start + 1); > + uint8_t *addr = block->host + (run_start << TARGET_PAGE_BITS); > + size_t chunk_size = (run_end - run_start) << TARGET_PAGE_BITS; > + > + if (qemu_madvise(addr, chunk_size, QEMU_MADV_DONTNEED) < 0) { > + error_report("migrate: madvise DONTNEED failed %p %ld: %s", > + addr, chunk_size, strerror(errno)); > + } can you use your ram_discard_page here, it keeps all the use of madvise together. > + > + run_start = find_next_zero_bit(bitmap, range, run_end + 1); > + } > + } > +} Dave > /* > * Callback from postcopy_each_ram_send_discard for each RAMBlock > * Note: At this point the 'unsentmap' is the processed bitmap combined > -- > 2.11.0 > > -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK