qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [PATCH v3 5/7] block: add throttle block filter driver


From: Kevin Wolf
Subject: Re: [Qemu-devel] [PATCH v3 5/7] block: add throttle block filter driver
Date: Thu, 3 Aug 2017 10:07:41 +0200
User-agent: Mutt/1.8.3 (2017-05-23)

Am 31.07.2017 um 11:54 hat Manos Pitsidianakis geschrieben:
> block/throttle.c uses existing I/O throttle infrastructure inside a
> block filter driver. I/O operations are intercepted in the filter's
> read/write coroutines, and referred to block/throttle-groups.c
> 
> The driver can be used with the syntax
> -drive driver=throttle,file.filename=foo.qcow2, \
>         limits.iops-total=...,throttle-group=bar
> 
> The configuration flags and their semantics are identical to the
> hardcoded throttling ones.
> 
> A node can be created referring to an existing group, and will overwrite
> its limits if any are specified, otherwise they are retained.
> 
> Signed-off-by: Manos Pitsidianakis <address@hidden>
> ---
>  block/Makefile.objs             |   1 +
>  block/throttle.c                | 395 
> ++++++++++++++++++++++++++++++++++++++++
>  include/qemu/throttle-options.h |   1 +
>  3 files changed, 397 insertions(+)
>  create mode 100644 block/throttle.c
> 
> diff --git a/block/Makefile.objs b/block/Makefile.objs
> index 2aaede4ae1..6eaf78a046 100644
> --- a/block/Makefile.objs
> +++ b/block/Makefile.objs
> @@ -25,6 +25,7 @@ block-obj-y += accounting.o dirty-bitmap.o
>  block-obj-y += write-threshold.o
>  block-obj-y += backup.o
>  block-obj-$(CONFIG_REPLICATION) += replication.o
> +block-obj-y += throttle.o
>  
>  block-obj-y += crypto.o
>  
> diff --git a/block/throttle.c b/block/throttle.c
> new file mode 100644
> index 0000000000..f3395462fb
> --- /dev/null
> +++ b/block/throttle.c
> @@ -0,0 +1,395 @@
> +/*
> + * QEMU block throttling filter driver infrastructure
> + *
> + * Copyright (c) 2017 Manos Pitsidianakis
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License as
> + * published by the Free Software Foundation; either version 2 or
> + * (at your option) version 3 of the License.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program; if not, see <http://www.gnu.org/licenses/>.
> + */
> +
> +#include "qemu/osdep.h"
> +#include "block/throttle-groups.h"
> +#include "qemu/throttle-options.h"
> +#include "qapi/error.h"
> +
> +#undef THROTTLE_OPT_PREFIX
> +#define THROTTLE_OPT_PREFIX "limits."
> +static QemuOptsList throttle_opts = {
> +    .name = "throttle",
> +    .head = QTAILQ_HEAD_INITIALIZER(throttle_opts.head),
> +    .desc = {
> +        THROTTLE_OPTS,
> +        {
> +            .name = QEMU_OPT_THROTTLE_GROUP_NAME,
> +            .type = QEMU_OPT_STRING,
> +            .help = "throttle group name",
> +        },
> +        { /* end of list */ }
> +    },
> +};
> +
> +/* Extract ThrottleConfig options. Assumes cfg is initialized and will be
> + * checked for validity.
> + */
> +static int throttle_extract_options(QemuOpts *opts, ThrottleConfig *cfg,
> +                                     Error **errp)
> +{
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_TOTAL)) {
> +        cfg->buckets[THROTTLE_BPS_TOTAL].avg =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_TOTAL,
> +                                0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_READ)) {
> +        cfg->buckets[THROTTLE_BPS_READ].avg  =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_READ,
> +                                0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_WRITE)) {
> +        cfg->buckets[THROTTLE_BPS_WRITE].avg =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_WRITE,
> +                                0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_TOTAL)) {
> +        cfg->buckets[THROTTLE_OPS_TOTAL].avg =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_IOPS_TOTAL,
> +                                0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_READ)) {
> +        cfg->buckets[THROTTLE_OPS_READ].avg =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_READ,
> +                                0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_WRITE)) {
> +        cfg->buckets[THROTTLE_OPS_WRITE].avg =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_IOPS_WRITE,
> +                                0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_TOTAL_MAX)) {
> +        cfg->buckets[THROTTLE_BPS_TOTAL].max =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_TOTAL_MAX, 0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_READ_MAX)) {
> +        cfg->buckets[THROTTLE_BPS_READ].max  =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_READ_MAX, 0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_BPS_WRITE_MAX)) {
> +        cfg->buckets[THROTTLE_BPS_WRITE].max =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_WRITE_MAX, 0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_TOTAL_MAX)) {
> +        cfg->buckets[THROTTLE_OPS_TOTAL].max =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_TOTAL_MAX, 0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_READ_MAX)) {
> +        cfg->buckets[THROTTLE_OPS_READ].max =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_READ_MAX, 0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_WRITE_MAX)) {
> +        cfg->buckets[THROTTLE_OPS_WRITE].max =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_WRITE_MAX, 0);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_BPS_TOTAL_MAX_LENGTH)) {
> +        if (qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_TOTAL_MAX_LENGTH, 1) > 
> UINT_MAX) {
> +            error_setg(errp, "%s value must be in the range [0, %u]",
> +                       THROTTLE_OPT_PREFIX QEMU_OPT_BPS_TOTAL_MAX_LENGTH,
> +                       UINT_MAX);
> +            return -1;
> +        }
> +        cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_TOTAL_MAX_LENGTH, 1);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_BPS_READ_MAX_LENGTH)) {
> +        if (qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_READ_MAX_LENGTH, 1) > UINT_MAX) 
> {
> +            error_setg(errp, "%s must be in the range [0, %u]",
> +                       THROTTLE_OPT_PREFIX QEMU_OPT_BPS_READ_MAX_LENGTH,
> +                       UINT_MAX);
> +            return -1;
> +        }
> +        cfg->buckets[THROTTLE_BPS_READ].burst_length  =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_READ_MAX_LENGTH, 1);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_BPS_WRITE_MAX_LENGTH)) {
> +        if (qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_WRITE_MAX_LENGTH, 1) > 
> UINT_MAX) {
> +            error_setg(errp, "%s must be in the range [0, %u]",
> +                       THROTTLE_OPT_PREFIX QEMU_OPT_BPS_WRITE_MAX_LENGTH,
> +                       UINT_MAX);
> +            return -1;
> +        }
> +        cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_BPS_WRITE_MAX_LENGTH, 1);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_IOPS_TOTAL_MAX_LENGTH)) {
> +        if (qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_TOTAL_MAX_LENGTH, 1) > 
> UINT_MAX) {
> +            error_setg(errp, "%s must be in the range [0, %u]",
> +                       THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_TOTAL_MAX_LENGTH,
> +                       UINT_MAX);
> +            return -1;
> +        }
> +        cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_TOTAL_MAX_LENGTH, 1);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_IOPS_READ_MAX_LENGTH)) {
> +        if (qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_READ_MAX_LENGTH, 1) > 
> UINT_MAX) {
> +            error_setg(errp, "%s must be in the range [0, %u]",
> +                       THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_READ_MAX_LENGTH,
> +                       UINT_MAX);
> +            return -1;
> +        }
> +        cfg->buckets[THROTTLE_OPS_READ].burst_length =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_READ_MAX_LENGTH, 1);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX 
> QEMU_OPT_IOPS_WRITE_MAX_LENGTH)) {
> +        if (qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_WRITE_MAX_LENGTH, 1) > 
> UINT_MAX) {
> +            error_setg(errp, "%s must be in the range [0, %u]",
> +                       THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_WRITE_MAX_LENGTH,
> +                       UINT_MAX);
> +            return -1;
> +        }
> +        cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX
> +                                QEMU_OPT_IOPS_WRITE_MAX_LENGTH, 1);
> +    }
> +    if (qemu_opt_get(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_SIZE)) {
> +        cfg->op_size =
> +            qemu_opt_get_number(opts, THROTTLE_OPT_PREFIX QEMU_OPT_IOPS_SIZE,
> +                                0);
> +    }
> +    return 0;
> +}

This function is very repetitive, but each block is long enough that
you have to look closely to review whether the right constants are used
everywhere.

Maybe this could become a bit more readable with a macro or two?

> +static int throttle_configure_tgm(BlockDriverState *bs,
> +                                  ThrottleGroupMember *tgm,
> +                                  QDict *options, Error **errp)
> +{
> +    int ret;
> +    ThrottleConfig cfg;
> +    const char *group_name = NULL;
> +    Error *local_err = NULL;
> +    QemuOpts *opts = qemu_opts_create(&throttle_opts, NULL, 0, &local_err);
> +    if (local_err) {
> +        error_propagate(errp, local_err);
> +        return -EINVAL;
> +    }

As Stefan said, qemu_opts_create() can't fail if you pass NULL for id,
so you can just use &error_abort.

> +
> +    qemu_opts_absorb_qdict(opts, options, &local_err);
> +    if (local_err) {
> +        error_propagate(errp, local_err);
> +        goto err;
> +    }
> +
> +    /* If no name is specified, an anonymous group will be created */
> +    group_name = qemu_opt_get(opts, QEMU_OPT_THROTTLE_GROUP_NAME);
> +
> +    /* Register membership to group with name group_name */
> +    throttle_group_register_tgm(tgm, group_name, bdrv_get_aio_context(bs));

The documentation of throttle_group_register_tgm() suggests that you
have to pass a string here, but group_name can be NULL if the option
wasn't given. Probably just means that the comment for that function
needs to be updated.

> +    /* Copy previous configuration */
> +    throttle_group_get_config(tgm, &cfg);
> +
> +    /* Change limits if user has specified them */
> +    if (throttle_extract_options(opts, &cfg, errp) ||
> +        !throttle_is_valid(&cfg, errp)) {
> +        throttle_group_unregister_tgm(tgm);
> +        goto err;
> +    }
> +    /* Update group configuration */
> +    throttle_group_config(tgm, &cfg);
> +
> +    ret = 0;
> +    goto fin;
> +
> +err:
> +    ret = -EINVAL;
> +fin:
> +    qemu_opts_del(opts);
> +    return ret;
> +}
> +
> +static int throttle_open(BlockDriverState *bs, QDict *options,
> +                            int flags, Error **errp)
> +{
> +    ThrottleGroupMember *tgm = bs->opaque;
> +
> +    bs->file = bdrv_open_child(NULL, options, "file",
> +                                    bs, &child_file, false, errp);

Indentation is off.

> +    if (!bs->file) {
> +        return -EINVAL;
> +    }
> +
> +    return throttle_configure_tgm(bs, tgm, options, errp);
> +}
> +
> +static void throttle_close(BlockDriverState *bs)
> +{
> +    ThrottleGroupMember *tgm = bs->opaque;
> +    throttle_group_unregister_tgm(tgm);
> +}
> +
> +
> +static int64_t throttle_getlength(BlockDriverState *bs)
> +{
> +    return bdrv_getlength(bs->file->bs);
> +}
> +
> +
> +static int coroutine_fn throttle_co_preadv(BlockDriverState *bs,
> +                                           uint64_t offset, uint64_t bytes,
> +                                           QEMUIOVector *qiov, int flags)
> +{
> +
> +    ThrottleGroupMember *tgm = bs->opaque;
> +    throttle_group_co_io_limits_intercept(tgm, bytes, false);
> +
> +    return bdrv_co_preadv(bs->file, offset, bytes, qiov, flags);
> +}
> +
> +static int coroutine_fn throttle_co_pwritev(BlockDriverState *bs,
> +                                            uint64_t offset, uint64_t bytes,
> +                                            QEMUIOVector *qiov, int flags)
> +{
> +    ThrottleGroupMember *tgm = bs->opaque;
> +    throttle_group_co_io_limits_intercept(tgm, bytes, true);
> +
> +    return bdrv_co_preadv(bs->file, offset, bytes, qiov, flags);
> +}

I think we want to set BlockDriver.supported_write_flags so that passing
down flags is actually of any use.

> +
> +static int coroutine_fn throttle_co_pwrite_zeroes(BlockDriverState *bs,
> +        int64_t offset, int bytes, BdrvRequestFlags flags)
> +{
> +    ThrottleGroupMember *tgm = bs->opaque;
> +    throttle_group_co_io_limits_intercept(tgm, bytes, true);
> +
> +    return bdrv_co_pwrite_zeroes(bs->file, offset, bytes, flags);
> +}

The same is true for BlockDriver.supported_zero_flags.

> +static int coroutine_fn throttle_co_pdiscard(BlockDriverState *bs,
> +        int64_t offset, int bytes)
> +{
> +    ThrottleGroupMember *tgm = bs->opaque;
> +    throttle_group_co_io_limits_intercept(tgm, bytes, true);
> +
> +    return bdrv_co_pdiscard(bs->file->bs, offset, bytes);
> +}
> +
> +static int throttle_co_flush(BlockDriverState *bs)
> +{
> +    return bdrv_co_flush(bs->file->bs);
> +}
> +
> +static void throttle_detach_aio_context(BlockDriverState *bs)
> +{
> +    ThrottleGroupMember *tgm = bs->opaque;
> +    throttle_group_detach_aio_context(tgm);
> +}
> +
> +static void throttle_attach_aio_context(BlockDriverState *bs,
> +                                    AioContext *new_context)

Indentation is off here...

> +{
> +    ThrottleGroupMember *tgm = bs->opaque;
> +    throttle_group_attach_aio_context(tgm, new_context);
> +}
> +
> +static int throttle_reopen_prepare(BDRVReopenState *reopen_state,
> +                                      BlockReopenQueue *queue, Error **errp)

...and here.

> +{
> +    ThrottleGroupMember *tgm = NULL;
> +
> +    assert(reopen_state != NULL);
> +    assert(reopen_state->bs != NULL);
> +
> +    reopen_state->opaque = g_new0(ThrottleGroupMember, 1);
> +    tgm = reopen_state->opaque;
> +
> +    return throttle_configure_tgm(reopen_state->bs, tgm, 
> reopen_state->options,
> +            errp);
> +}
> +
> +static void throttle_reopen_commit(BDRVReopenState *state)
> +{
> +    ThrottleGroupMember *tgm = state->bs->opaque;
> +
> +    throttle_group_unregister_tgm(tgm);
> +    g_free(state->bs->opaque);
> +    state->bs->opaque = state->opaque;
> +    state->opaque = NULL;
> +}
> +
> +static void throttle_reopen_abort(BDRVReopenState *state)
> +{
> +    ThrottleGroupMember *tgm = state->opaque;
> +
> +    throttle_group_unregister_tgm(tgm);
> +    g_free(state->opaque);
> +    state->opaque = NULL;
> +}
> +
> +static bool throttle_recurse_is_first_non_filter(BlockDriverState *bs,
> +                                                 BlockDriverState *candidate)
> +{
> +    return bdrv_recurse_is_first_non_filter(bs->file->bs, candidate);
> +}
> +
> +static BlockDriver bdrv_throttle = {
> +    .format_name                        =   "throttle",
> +    .protocol_name                      =   "throttle",
> +    .instance_size                      =   sizeof(ThrottleGroupMember),
> +
> +    .bdrv_file_open                     =   throttle_open,
> +    .bdrv_close                         =   throttle_close,
> +    .bdrv_co_flush                      =   throttle_co_flush,
> +
> +    .bdrv_child_perm                    =   bdrv_filter_default_perms,
> +
> +    .bdrv_getlength                     =   throttle_getlength,
> +
> +    .bdrv_co_preadv                     =   throttle_co_preadv,
> +    .bdrv_co_pwritev                    =   throttle_co_pwritev,
> +
> +    .bdrv_co_pwrite_zeroes              =   throttle_co_pwrite_zeroes,
> +    .bdrv_co_pdiscard                   =   throttle_co_pdiscard,
> +
> +    .bdrv_recurse_is_first_non_filter   =   
> throttle_recurse_is_first_non_filter,
> +
> +    .bdrv_attach_aio_context            =   throttle_attach_aio_context,
> +    .bdrv_detach_aio_context            =   throttle_detach_aio_context,
> +
> +    .bdrv_reopen_prepare                =   throttle_reopen_prepare,
> +    .bdrv_reopen_commit                 =   throttle_reopen_commit,
> +    .bdrv_reopen_abort                  =   throttle_reopen_abort,
> +
> +    .is_filter                          =   true,
> +};

What about .bdrv_co_get_block_status?

> +static void bdrv_throttle_init(void)
> +{
> +    bdrv_register(&bdrv_throttle);
> +}
> +
> +block_init(bdrv_throttle_init);
> diff --git a/include/qemu/throttle-options.h b/include/qemu/throttle-options.h
> index 182b7896e1..3528a8f4a2 100644
> --- a/include/qemu/throttle-options.h
> +++ b/include/qemu/throttle-options.h
> @@ -29,6 +29,7 @@
>  #define QEMU_OPT_BPS_WRITE_MAX "bps-write-max"
>  #define QEMU_OPT_BPS_WRITE_MAX_LENGTH "bps-write-max-length"
>  #define QEMU_OPT_IOPS_SIZE "iops-size"
> +#define QEMU_OPT_THROTTLE_GROUP_NAME "throttle-group"
>  
>  #define THROTTLE_OPT_PREFIX "throttling."
>  #define THROTTLE_OPTS \

Kevin



reply via email to

[Prev in Thread] Current Thread [Next in Thread]