From patchwork Wed Apr 11 18:54:14 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Max Reitz X-Patchwork-Id: 897353 Return-Path: X-Original-To: incoming@patchwork.ozlabs.org Delivered-To: patchwork-incoming@bilbo.ozlabs.org Authentication-Results: ozlabs.org; spf=pass (mailfrom) smtp.mailfrom=nongnu.org (client-ip=2001:4830:134:3::11; helo=lists.gnu.org; envelope-from=qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org; receiver=) Authentication-Results: ozlabs.org; dmarc=fail (p=none dis=none) header.from=redhat.com Received: from lists.gnu.org (lists.gnu.org [IPv6:2001:4830:134:3::11]) (using TLSv1 with cipher AES256-SHA (256/256 bits)) (No client certificate requested) by ozlabs.org (Postfix) with ESMTPS id 40Ltf06jqwz9s1t for ; Thu, 12 Apr 2018 05:00:24 +1000 (AEST) Received: from localhost ([::1]:48478 helo=lists.gnu.org) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1f6Kyk-0008C6-Qe for incoming@patchwork.ozlabs.org; Wed, 11 Apr 2018 15:00:22 -0400 Received: from eggs.gnu.org ([2001:4830:134:3::10]:36127) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1f6KtK-00035K-2B for qemu-devel@nongnu.org; Wed, 11 Apr 2018 14:54:49 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1f6KtI-0005va-JT for qemu-devel@nongnu.org; Wed, 11 Apr 2018 14:54:46 -0400 Received: from mx3-rdu2.redhat.com ([66.187.233.73]:39172 helo=mx1.redhat.com) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1f6KtE-0005r6-By; Wed, 11 Apr 2018 14:54:40 -0400 Received: from smtp.corp.redhat.com (int-mx06.intmail.prod.int.rdu2.redhat.com [10.11.54.6]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id DD4AF8010F67; Wed, 11 Apr 2018 18:54:39 +0000 (UTC) Received: from localhost (ovpn-204-29.brq.redhat.com [10.40.204.29]) by smtp.corp.redhat.com (Postfix) with ESMTPS id 562B0215CDC6; Wed, 11 Apr 2018 18:54:39 +0000 (UTC) From: Max Reitz To: qemu-block@nongnu.org Date: Wed, 11 Apr 2018 20:54:14 +0200 Message-Id: <20180411185425.2461-3-mreitz@redhat.com> In-Reply-To: <20180411185425.2461-1-mreitz@redhat.com> References: <20180411185425.2461-1-mreitz@redhat.com> X-Scanned-By: MIMEDefang 2.78 on 10.11.54.6 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.11.55.8]); Wed, 11 Apr 2018 18:54:39 +0000 (UTC) X-Greylist: inspected by milter-greylist-4.5.16 (mx1.redhat.com [10.11.55.8]); Wed, 11 Apr 2018 18:54:39 +0000 (UTC) for IP:'10.11.54.6' DOMAIN:'int-mx06.intmail.prod.int.rdu2.redhat.com' HELO:'smtp.corp.redhat.com' FROM:'mreitz@redhat.com' RCPT:'' X-detected-operating-system: by eggs.gnu.org: GNU/Linux 2.2.x-3.x [generic] [fuzzy] X-Received-From: 66.187.233.73 Subject: [Qemu-devel] [PATCH v4 02/13] block/mirror: Convert to coroutines X-BeenThere: qemu-devel@nongnu.org X-Mailman-Version: 2.1.21 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Kevin Wolf , John Snow , qemu-devel@nongnu.org, Stefan Hajnoczi , Max Reitz Errors-To: qemu-devel-bounces+incoming=patchwork.ozlabs.org@nongnu.org Sender: "Qemu-devel" In order to talk to the source BDS (and maybe in the future to the target BDS as well) directly, we need to convert our existing AIO requests into coroutine I/O requests. Signed-off-by: Max Reitz Reviewed-by: Fam Zheng --- block/mirror.c | 152 ++++++++++++++++++++++++++++++++++----------------------- 1 file changed, 90 insertions(+), 62 deletions(-) diff --git a/block/mirror.c b/block/mirror.c index 1718571766..bd8ee7d92c 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -80,6 +80,10 @@ typedef struct MirrorOp { QEMUIOVector qiov; int64_t offset; uint64_t bytes; + + /* The pointee is set by mirror_co_read(), mirror_co_zero(), and + * mirror_co_discard() before yielding for the first time */ + int64_t *bytes_handled; } MirrorOp; typedef enum MirrorMethod { @@ -101,7 +105,7 @@ static BlockErrorAction mirror_error_action(MirrorBlockJob *s, bool read, } } -static void mirror_iteration_done(MirrorOp *op, int ret) +static void coroutine_fn mirror_iteration_done(MirrorOp *op, int ret) { MirrorBlockJob *s = op->s; struct iovec *iov; @@ -138,9 +142,8 @@ static void mirror_iteration_done(MirrorOp *op, int ret) } } -static void mirror_write_complete(void *opaque, int ret) +static void coroutine_fn mirror_write_complete(MirrorOp *op, int ret) { - MirrorOp *op = opaque; MirrorBlockJob *s = op->s; aio_context_acquire(blk_get_aio_context(s->common.blk)); @@ -157,9 +160,8 @@ static void mirror_write_complete(void *opaque, int ret) aio_context_release(blk_get_aio_context(s->common.blk)); } -static void mirror_read_complete(void *opaque, int ret) +static void coroutine_fn mirror_read_complete(MirrorOp *op, int ret) { - MirrorOp *op = opaque; MirrorBlockJob *s = op->s; aio_context_acquire(blk_get_aio_context(s->common.blk)); @@ -174,8 +176,9 @@ static void mirror_read_complete(void *opaque, int ret) mirror_iteration_done(op, ret); } else { - blk_aio_pwritev(s->target, op->offset, &op->qiov, - 0, mirror_write_complete, op); + ret = blk_co_pwritev(s->target, op->offset, + op->qiov.size, &op->qiov, 0); + mirror_write_complete(op, ret); } aio_context_release(blk_get_aio_context(s->common.blk)); } @@ -232,60 +235,57 @@ static inline void mirror_wait_for_io(MirrorBlockJob *s) s->waiting_for_io = false; } -/* Submit async read while handling COW. - * Returns: The number of bytes copied after and including offset, - * excluding any bytes copied prior to offset due to alignment. - * This will be @bytes if no alignment is necessary, or - * (new_end - offset) if tail is rounded up or down due to - * alignment or buffer limit. +/* Perform a mirror copy operation. + * + * *op->bytes_handled is set to the number of bytes copied after and + * including offset, excluding any bytes copied prior to offset due + * to alignment. This will be op->bytes if no alignment is necessary, + * or (new_end - op->offset) if the tail is rounded up or down due to + * alignment or buffer limit. */ -static uint64_t mirror_do_read(MirrorBlockJob *s, int64_t offset, - uint64_t bytes) +static void coroutine_fn mirror_co_read(void *opaque) { + MirrorOp *op = opaque; + MirrorBlockJob *s = op->s; BlockBackend *source = s->common.blk; int nb_chunks; uint64_t ret; - MirrorOp *op; uint64_t max_bytes; max_bytes = s->granularity * s->max_iov; /* We can only handle as much as buf_size at a time. */ - bytes = MIN(s->buf_size, MIN(max_bytes, bytes)); - assert(bytes); - assert(bytes < BDRV_REQUEST_MAX_BYTES); - ret = bytes; + op->bytes = MIN(s->buf_size, MIN(max_bytes, op->bytes)); + assert(op->bytes); + assert(op->bytes < BDRV_REQUEST_MAX_BYTES); + *op->bytes_handled = op->bytes; if (s->cow_bitmap) { - ret += mirror_cow_align(s, &offset, &bytes); + *op->bytes_handled += mirror_cow_align(s, &op->offset, &op->bytes); } - assert(bytes <= s->buf_size); + /* Cannot exceed BDRV_REQUEST_MAX_BYTES + INT_MAX */ + assert(*op->bytes_handled <= UINT_MAX); + assert(op->bytes <= s->buf_size); /* The offset is granularity-aligned because: * 1) Caller passes in aligned values; * 2) mirror_cow_align is used only when target cluster is larger. */ - assert(QEMU_IS_ALIGNED(offset, s->granularity)); + assert(QEMU_IS_ALIGNED(op->offset, s->granularity)); /* The range is sector-aligned, since bdrv_getlength() rounds up. */ - assert(QEMU_IS_ALIGNED(bytes, BDRV_SECTOR_SIZE)); - nb_chunks = DIV_ROUND_UP(bytes, s->granularity); + assert(QEMU_IS_ALIGNED(op->bytes, BDRV_SECTOR_SIZE)); + nb_chunks = DIV_ROUND_UP(op->bytes, s->granularity); while (s->buf_free_count < nb_chunks) { - trace_mirror_yield_in_flight(s, offset, s->in_flight); + trace_mirror_yield_in_flight(s, op->offset, s->in_flight); mirror_wait_for_io(s); } - /* Allocate a MirrorOp that is used as an AIO callback. */ - op = g_new(MirrorOp, 1); - op->s = s; - op->offset = offset; - op->bytes = bytes; - /* Now make a QEMUIOVector taking enough granularity-sized chunks * from s->buf_free. */ qemu_iovec_init(&op->qiov, nb_chunks); while (nb_chunks-- > 0) { MirrorBuffer *buf = QSIMPLEQ_FIRST(&s->buf_free); - size_t remaining = bytes - op->qiov.size; + size_t remaining = op->bytes - op->qiov.size; QSIMPLEQ_REMOVE_HEAD(&s->buf_free, next); s->buf_free_count--; @@ -294,53 +294,81 @@ static uint64_t mirror_do_read(MirrorBlockJob *s, int64_t offset, /* Copy the dirty cluster. */ s->in_flight++; - s->bytes_in_flight += bytes; - trace_mirror_one_iteration(s, offset, bytes); + s->bytes_in_flight += op->bytes; + trace_mirror_one_iteration(s, op->offset, op->bytes); - blk_aio_preadv(source, offset, &op->qiov, 0, mirror_read_complete, op); - return ret; + ret = blk_co_preadv(source, op->offset, op->bytes, &op->qiov, 0); + mirror_read_complete(op, ret); } -static void mirror_do_zero_or_discard(MirrorBlockJob *s, - int64_t offset, - uint64_t bytes, - bool is_discard) +static void coroutine_fn mirror_co_zero(void *opaque) { - MirrorOp *op; + MirrorOp *op = opaque; + int ret; - /* Allocate a MirrorOp that is used as an AIO callback. The qiov is zeroed - * so the freeing in mirror_iteration_done is nop. */ - op = g_new0(MirrorOp, 1); - op->s = s; - op->offset = offset; - op->bytes = bytes; + op->s->in_flight++; + op->s->bytes_in_flight += op->bytes; + *op->bytes_handled = op->bytes; - s->in_flight++; - s->bytes_in_flight += bytes; - if (is_discard) { - blk_aio_pdiscard(s->target, offset, - op->bytes, mirror_write_complete, op); - } else { - blk_aio_pwrite_zeroes(s->target, offset, - op->bytes, s->unmap ? BDRV_REQ_MAY_UNMAP : 0, - mirror_write_complete, op); - } + ret = blk_co_pwrite_zeroes(op->s->target, op->offset, op->bytes, + op->s->unmap ? BDRV_REQ_MAY_UNMAP : 0); + mirror_write_complete(op, ret); +} + +static void coroutine_fn mirror_co_discard(void *opaque) +{ + MirrorOp *op = opaque; + int ret; + + op->s->in_flight++; + op->s->bytes_in_flight += op->bytes; + *op->bytes_handled = op->bytes; + + ret = blk_co_pdiscard(op->s->target, op->offset, op->bytes); + mirror_write_complete(op, ret); } static unsigned mirror_perform(MirrorBlockJob *s, int64_t offset, unsigned bytes, MirrorMethod mirror_method) { + MirrorOp *op; + Coroutine *co; + int64_t bytes_handled = -1; + + op = g_new(MirrorOp, 1); + *op = (MirrorOp){ + .s = s, + .offset = offset, + .bytes = bytes, + .bytes_handled = &bytes_handled, + }; + switch (mirror_method) { case MIRROR_METHOD_COPY: - return mirror_do_read(s, offset, bytes); + co = qemu_coroutine_create(mirror_co_read, op); + break; case MIRROR_METHOD_ZERO: + co = qemu_coroutine_create(mirror_co_zero, op); + break; case MIRROR_METHOD_DISCARD: - mirror_do_zero_or_discard(s, offset, bytes, - mirror_method == MIRROR_METHOD_DISCARD); - return bytes; + co = qemu_coroutine_create(mirror_co_discard, op); + break; default: abort(); } + + qemu_coroutine_enter(co); + /* At this point, ownership of op has been moved to the coroutine + * and the object may already be freed */ + + /* Assert that this value has been set */ + assert(bytes_handled >= 0); + + /* Same assertion as in mirror_co_read() (and for mirror_co_read() + * and mirror_co_discard(), bytes_handled == op->bytes, which + * is the @bytes parameter given to this function) */ + assert(bytes_handled <= UINT_MAX); + return bytes_handled; } static uint64_t coroutine_fn mirror_iteration(MirrorBlockJob *s)