qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Denis V. Lunev" <den@openvz.org>
To: Roman Kagan <rkagan@parallels.com>, Kevin Wolf <kwolf@redhat.com>,
	qemu-devel@nongnu.org, Stefan Hajnoczi <stefanha@redhat.com>
Subject: Re: [Qemu-devel] [PATCH 17/19] block/parallels: delay writing to BAT till bdrv_co_flush_to_os
Date: Wed, 14 Jan 2015 17:29:44 +0300	[thread overview]
Message-ID: <54B67D58.3060806@openvz.org> (raw)
In-Reply-To: <20150114133447.GC27418@rkaganb.sw.ru>

On 14/01/15 16:34, Roman Kagan wrote:
> On Wed, Jan 14, 2015 at 04:08:50PM +0300, Denis V. Lunev wrote:
>> On 14/01/15 16:03, Roman Kagan wrote:
>>> On Tue, Dec 30, 2014 at 01:07:10PM +0300, Denis V. Lunev wrote:
>>>> +static int cache_bat(BlockDriverState *bs, uint32_t idx, uint32_t new_data_off)
>>>> +{
>>>> +    int ret, i, off, cache_off;
>>>> +    int64_t first_idx, last_idx;
>>>> +    BDRVParallelsState *s = bs->opaque;
>>>> +    uint32_t *cache = s->bat_cache;
>>>> +
>>>> +    off = bat_offset(idx);
>>>> +    cache_off = (off / s->bat_cache_size) * s->bat_cache_size;
>>>> +
>>>> +    if (s->bat_cache_off != -1 && s->bat_cache_off != cache_off) {
>>>> +        ret = write_bat_cache(bs);
>>>> +        if (ret < 0) {
>>>> +            return ret;
>>>> +        }
>>>> +    }
>>>> +
>>>> +    first_idx = idx - (off - cache_off) / sizeof(uint32_t);
>>>> +    last_idx = first_idx + s->bat_cache_size / sizeof(uint32_t);
>>>> +    if (first_idx < 0) {
>>>> +        memcpy(s->bat_cache, &s->ph, sizeof(s->ph));
>>>> +        first_idx = 0;
>>>> +        cache = s->bat_cache + sizeof(s->ph) / sizeof(uint32_t);
>>>> +    }
>>>> +
>>>> +    if (last_idx > s->bat_size) {
>>>> +        memset(cache + s->bat_size - first_idx, 0,
>>>> +               sizeof(uint32_t) * (last_idx - s->bat_size));
>>>> +    }
>>>> +
>>>> +    for (i = 0; i < last_idx - first_idx; i++) {
>>>> +        cache[i] = cpu_to_le32(s->bat[first_idx + i]);
>>>> +    }
>>> You're re-populating the bat_cache on every bat update.  Why?
>>>
>>> Shouldn't this be done only with empty cache, i.e. under
>>> if (s->bat_cache_off == -1)?
>> reasonable, but condition should be a bit different, namely
>>
>> if (s->bat_cache_off != -1 && s->bat_cache_off != cache_off) {
> No, this is the condition to flush the cache which you already do.
>
> Then, either upon flushing it or on the first entry into cache_bat(),
> the cache is empty which is indicated by ->bat_cache_off == -1, at which
> point you need to populate it from ->bat.

you are wrong. BAT cache is a single page of the whole BAT
which can be more than several MBs. This page should be
repopulated when the off is changed.

  reply	other threads:[~2015-01-14 14:29 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-12-30  9:28 [Qemu-devel] [PATCH v2 0/19] write/create for Parallels images with reasonable performance Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 01/19] iotests, parallels: quote TEST_IMG in 076 test to be path-safe Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 02/19] block/parallels: rename parallels_header to ParallelsHeader Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 03/19] block/parallels: switch to bdrv_read Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 04/19] block/parallels: read up to cluster end in one go Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 05/19] block/parallels: add get_block_status Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 06/19] block/parallels: provide _co_readv routine for parallels format driver Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 07/19] block/parallels: replace magic constants 4, 64 with proper sizeofs Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 08/19] block/parallels: _co_writev callback for Parallels format Denis V. Lunev
2014-12-30  9:28 ` [Qemu-devel] [PATCH 09/19] iotests, parallels: test for write into Parallels image Denis V. Lunev
2014-12-30 10:07 ` [Qemu-devel] [PATCH 10/19] block/parallels: support parallels image creation Denis V. Lunev
2014-12-30 10:07   ` [Qemu-devel] [PATCH 11/19] iotests, parallels: test for newly created parallels image via qemu-img Denis V. Lunev
2014-12-30 10:07   ` [Qemu-devel] [PATCH 12/19] parallels: change copyright information in the image header Denis V. Lunev
2014-12-30 10:07   ` [Qemu-devel] [PATCH 13/19] block/parallels: store ParallelsHeader to the BDRVParallelsState Denis V. Lunev
2015-01-13 14:13     ` Roman Kagan
2014-12-30 10:07   ` [Qemu-devel] [PATCH 14/19] block/parallels: create catalog_offset helper Denis V. Lunev
2015-01-13 14:14     ` Roman Kagan
2014-12-30 10:07   ` [Qemu-devel] [PATCH 15/19] block/parallels: rename catalog_ names to bat_ Denis V. Lunev
2015-01-13 14:15     ` Roman Kagan
2014-12-30 10:07   ` [Qemu-devel] [PATCH 16/19] block/parallels: no need to flush on each block allocation table update Denis V. Lunev
2015-01-13 14:50     ` Roman Kagan
2015-01-13 15:17       ` Denis V. Lunev
2015-01-13 20:16         ` Denis V. Lunev
2015-01-14 12:01           ` Roman Kagan
2015-01-14 12:30             ` Denis V. Lunev
2014-12-30 10:07   ` [Qemu-devel] [PATCH 17/19] block/parallels: delay writing to BAT till bdrv_co_flush_to_os Denis V. Lunev
2015-01-14 13:03     ` Roman Kagan
2015-01-14 13:08       ` Denis V. Lunev
2015-01-14 13:34         ` Roman Kagan
2015-01-14 14:29           ` Denis V. Lunev [this message]
2015-01-14 16:44             ` Denis V. Lunev
2014-12-30 10:07   ` [Qemu-devel] [PATCH 18/19] block/parallels: add prealloc-mode and prealloc-size open paramemets Denis V. Lunev
2015-01-14 14:26     ` Roman Kagan
2015-01-14 14:31       ` Denis V. Lunev
2015-01-14 17:22         ` Roman Kagan
2014-12-30 10:07   ` [Qemu-devel] [PATCH 19/19] block/parallels: optimize linear image expansion Denis V. Lunev
2015-01-14 17:56     ` Roman Kagan
2015-01-14 17:57       ` Denis V. Lunev

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=54B67D58.3060806@openvz.org \
    --to=den@openvz.org \
    --cc=kwolf@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=rkagan@parallels.com \
    --cc=stefanha@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).