diff mbox series

[v5,06/12] migration/dirtyrate: Record hash results for each sampled page

Message ID 1598260480-64862-7-git-send-email-zhengchuan@huawei.com (mailing list archive)
State New, archived
Headers show
Series *** A Method for evaluating dirty page rate *** | expand

Commit Message

Zheng Chuan Aug. 24, 2020, 9:14 a.m. UTC
Record hash results for each sampled page, crc32 is taken to calculate
hash results for each sampled 4K-page.

Signed-off-by: Chuan Zheng <zhengchuan@huawei.com>
Signed-off-by: YanYing Zhuang <ann.zhuangyanying@huawei.com>
---
 migration/dirtyrate.c | 136 ++++++++++++++++++++++++++++++++++++++++++++++++++
 migration/dirtyrate.h |  15 ++++++
 2 files changed, 151 insertions(+)

Comments

David Edmondson Aug. 26, 2020, 9:56 a.m. UTC | #1
On Monday, 2020-08-24 at 17:14:34 +08, Chuan Zheng wrote:

> Record hash results for each sampled page, crc32 is taken to calculate
> hash results for each sampled 4K-page.
>
> Signed-off-by: Chuan Zheng <zhengchuan@huawei.com>
> Signed-off-by: YanYing Zhuang <ann.zhuangyanying@huawei.com>
> ---
>  migration/dirtyrate.c | 136 ++++++++++++++++++++++++++++++++++++++++++++++++++
>  migration/dirtyrate.h |  15 ++++++
>  2 files changed, 151 insertions(+)
>
> diff --git a/migration/dirtyrate.c b/migration/dirtyrate.c
> index f6a94d8..66de426 100644
> --- a/migration/dirtyrate.c
> +++ b/migration/dirtyrate.c
> @@ -10,6 +10,7 @@
>   * See the COPYING file in the top-level directory.
>   */
>  
> +#include <zlib.h>
>  #include "qemu/osdep.h"
>  #include "qapi/error.h"
>  #include "crypto/hash.h"
> @@ -66,6 +67,141 @@ static void update_dirtyrate(uint64_t msec)
>      DirtyStat.dirty_rate = dirtyrate;
>  }
>  
> +/*
> + * get hash result for the sampled memory with length of 4K byte in ramblock,
> + * which starts from ramblock base address.
> + */
> +static uint32_t get_ramblock_vfn_hash(struct RamblockDirtyInfo *info,
> +                                      uint64_t vfn)
> +{
> +    struct iovec iov_array;

There's no need for an iovec now that crc32() is being used.

> +    uint32_t crc;
> +
> +    iov_array.iov_base = info->ramblock_addr +
> +                         vfn * DIRTYRATE_SAMPLE_PAGE_SIZE;
> +    iov_array.iov_len = DIRTYRATE_SAMPLE_PAGE_SIZE;
> +
> +    crc = crc32(0, iov_array.iov_base, iov_array.iov_len);
> +
> +    return crc;
> +}
> +
> +static int save_ramblock_hash(struct RamblockDirtyInfo *info)
> +{
> +    unsigned int sample_pages_count;
> +    int i;
> +    int ret = -1;

There's no need to initialise "ret".

> +    GRand *rand = g_rand_new();

Calling g_rand_new() when the result may not be used (because of the
various conditions immediately below) seems as though it might waste
entropy. Could this be pushed down just above the loop? It would even
get rid of the gotos ;-)

> +    sample_pages_count = info->sample_pages_count;
> +
> +    /* ramblock size less than one page, return success to skip this ramblock */
> +    if (unlikely(info->ramblock_pages == 0 || sample_pages_count == 0)) {
> +        ret = 0;
> +        goto out;
> +    }
> +
> +    info->hash_result = g_try_malloc0_n(sample_pages_count,
> +                                        sizeof(uint32_t));
> +    if (!info->hash_result) {
> +        ret = -1;
> +        goto out;
> +    }
> +
> +    info->sample_page_vfn = g_try_malloc0_n(sample_pages_count,
> +                                            sizeof(uint64_t));
> +    if (!info->sample_page_vfn) {
> +        g_free(info->hash_result);
> +        ret = -1;
> +        goto out;
> +    }
> +
> +    for (i = 0; i < sample_pages_count; i++) {
> +        info->sample_page_vfn[i] = g_rand_int_range(rand, 0,
> +                                                    info->ramblock_pages - 1);
> +        info->hash_result[i] = get_ramblock_vfn_hash(info,
> +                                                     info->sample_page_vfn[i]);
> +    }
> +    ret = 0;
> +
> +out:
> +    g_rand_free(rand);
> +    return ret;
> +}
> +
> +static void get_ramblock_dirty_info(RAMBlock *block,
> +                                    struct RamblockDirtyInfo *info,
> +                                    struct DirtyRateConfig *config)
> +{
> +    uint64_t sample_pages_per_gigabytes = config->sample_pages_per_gigabytes;
> +
> +    /* Right shift 30 bits to calc block size in GB */
> +    info->sample_pages_count = (qemu_ram_get_used_length(block) *
> +                                sample_pages_per_gigabytes) >>
> +                                DIRTYRATE_PAGE_SHIFT_GB;

Doing the calculation this way around seems odd. Why was it not:

    info->sample_pages_count = (qemu_ram_get_used_length(block) >>
                                DIRTYRATE_PAGE_SHIFT_GB) *
                                sample_pages_per_gigabytes;

?

> +
> +    /* Right shift 12 bits to calc page count in 4KB */
> +    info->ramblock_pages = qemu_ram_get_used_length(block) >>
> +                           DIRTYRATE_PAGE_SHIFT_KB;
> +    info->ramblock_addr = qemu_ram_get_host_addr(block);
> +    strcpy(info->idstr, qemu_ram_get_idstr(block));
> +}
> +
> +static struct RamblockDirtyInfo *
> +alloc_ramblock_dirty_info(int *block_index,
> +                          struct RamblockDirtyInfo *block_dinfo)
> +{
> +    struct RamblockDirtyInfo *info = NULL;
> +    int index = *block_index;
> +
> +    if (!block_dinfo) {
> +        index = 0;
> +        block_dinfo = g_try_new(struct RamblockDirtyInfo, 1);
> +    } else {
> +        index++;
> +        block_dinfo = g_try_realloc(block_dinfo, (index + 1) *
> +                                    sizeof(struct RamblockDirtyInfo));

g_try_renew() instead of g_try_realloc()?

> +    }
> +    if (!block_dinfo) {
> +        return NULL;
> +    }
> +
> +    info = &block_dinfo[index];
> +    *block_index = index;
> +    memset(info, 0, sizeof(struct RamblockDirtyInfo));
> +
> +    return block_dinfo;
> +}
> +
> +static int record_ramblock_hash_info(struct RamblockDirtyInfo **block_dinfo,
> +                                     struct DirtyRateConfig config,
> +                                     int *block_index)
> +{
> +    struct RamblockDirtyInfo *info = NULL;
> +    struct RamblockDirtyInfo *dinfo = NULL;
> +    RAMBlock *block = NULL;

There's no need to initialise "info" or "block".

The declaration of "info" could be pushed into the loop.

> +    int index = 0;
> +
> +    RAMBLOCK_FOREACH_MIGRATABLE(block) {
> +        dinfo = alloc_ramblock_dirty_info(&index, dinfo);
> +        if (dinfo == NULL) {
> +            return -1;
> +        }
> +        info = &dinfo[index];
> +        get_ramblock_dirty_info(block, info, &config);
> +        if (save_ramblock_hash(info) < 0) {
> +            *block_dinfo = dinfo;
> +            *block_index = index;
> +            return -1;
> +        }
> +    }
> +
> +    *block_dinfo = dinfo;
> +    *block_index = index;
> +
> +    return 0;
> +}
> +
>  static void calculate_dirtyrate(struct DirtyRateConfig config)
>  {
>      /* todo */
> diff --git a/migration/dirtyrate.h b/migration/dirtyrate.h
> index 9db269d..5050add 100644
> --- a/migration/dirtyrate.h
> +++ b/migration/dirtyrate.h
> @@ -24,6 +24,21 @@
>   */
>  #define RAMBLOCK_INFO_MAX_LEN                     256
>  
> +/*
> + * Sample page size 4K as default.
> + */
> +#define DIRTYRATE_SAMPLE_PAGE_SIZE                4096
> +
> +/*
> + * Sample page size 4K shift
> + */
> +#define DIRTYRATE_PAGE_SHIFT_KB                   12
> +
> +/*
> + * Sample page size 1G shift
> + */
> +#define DIRTYRATE_PAGE_SHIFT_GB                   30
> +
>  /* Take 1s as default for calculation duration */
>  #define DEFAULT_FETCH_DIRTYRATE_TIME_SEC          1
>  
> -- 
> 1.8.3.1

dme.
Dr. David Alan Gilbert Aug. 26, 2020, 12:30 p.m. UTC | #2
* David Edmondson (dme@dme.org) wrote:
> On Monday, 2020-08-24 at 17:14:34 +08, Chuan Zheng wrote:
> 
> > Record hash results for each sampled page, crc32 is taken to calculate
> > hash results for each sampled 4K-page.
> >
> > Signed-off-by: Chuan Zheng <zhengchuan@huawei.com>
> > Signed-off-by: YanYing Zhuang <ann.zhuangyanying@huawei.com>
> > ---
> >  migration/dirtyrate.c | 136 ++++++++++++++++++++++++++++++++++++++++++++++++++
> >  migration/dirtyrate.h |  15 ++++++
> >  2 files changed, 151 insertions(+)
> >
> > diff --git a/migration/dirtyrate.c b/migration/dirtyrate.c
> > index f6a94d8..66de426 100644
> > --- a/migration/dirtyrate.c
> > +++ b/migration/dirtyrate.c
> > @@ -10,6 +10,7 @@
> >   * See the COPYING file in the top-level directory.
> >   */
> >  
> > +#include <zlib.h>
> >  #include "qemu/osdep.h"
> >  #include "qapi/error.h"
> >  #include "crypto/hash.h"
> > @@ -66,6 +67,141 @@ static void update_dirtyrate(uint64_t msec)
> >      DirtyStat.dirty_rate = dirtyrate;
> >  }
> >  
> > +/*
> > + * get hash result for the sampled memory with length of 4K byte in ramblock,
> > + * which starts from ramblock base address.
> > + */
> > +static uint32_t get_ramblock_vfn_hash(struct RamblockDirtyInfo *info,
> > +                                      uint64_t vfn)
> > +{
> > +    struct iovec iov_array;
> 
> There's no need for an iovec now that crc32() is being used.
> 
> > +    uint32_t crc;
> > +
> > +    iov_array.iov_base = info->ramblock_addr +
> > +                         vfn * DIRTYRATE_SAMPLE_PAGE_SIZE;
> > +    iov_array.iov_len = DIRTYRATE_SAMPLE_PAGE_SIZE;
> > +
> > +    crc = crc32(0, iov_array.iov_base, iov_array.iov_len);
> > +
> > +    return crc;
> > +}
> > +
> > +static int save_ramblock_hash(struct RamblockDirtyInfo *info)
> > +{
> > +    unsigned int sample_pages_count;
> > +    int i;
> > +    int ret = -1;
> 
> There's no need to initialise "ret".
> 
> > +    GRand *rand = g_rand_new();
> 
> Calling g_rand_new() when the result may not be used (because of the
> various conditions immediately below) seems as though it might waste
> entropy. Could this be pushed down just above the loop? It would even
> get rid of the gotos ;-)
> 
> > +    sample_pages_count = info->sample_pages_count;
> > +
> > +    /* ramblock size less than one page, return success to skip this ramblock */
> > +    if (unlikely(info->ramblock_pages == 0 || sample_pages_count == 0)) {
> > +        ret = 0;
> > +        goto out;
> > +    }
> > +
> > +    info->hash_result = g_try_malloc0_n(sample_pages_count,
> > +                                        sizeof(uint32_t));
> > +    if (!info->hash_result) {
> > +        ret = -1;
> > +        goto out;
> > +    }
> > +
> > +    info->sample_page_vfn = g_try_malloc0_n(sample_pages_count,
> > +                                            sizeof(uint64_t));
> > +    if (!info->sample_page_vfn) {
> > +        g_free(info->hash_result);
> > +        ret = -1;
> > +        goto out;
> > +    }
> > +
> > +    for (i = 0; i < sample_pages_count; i++) {
> > +        info->sample_page_vfn[i] = g_rand_int_range(rand, 0,
> > +                                                    info->ramblock_pages - 1);
> > +        info->hash_result[i] = get_ramblock_vfn_hash(info,
> > +                                                     info->sample_page_vfn[i]);
> > +    }
> > +    ret = 0;
> > +
> > +out:
> > +    g_rand_free(rand);
> > +    return ret;
> > +}
> > +
> > +static void get_ramblock_dirty_info(RAMBlock *block,
> > +                                    struct RamblockDirtyInfo *info,
> > +                                    struct DirtyRateConfig *config)
> > +{
> > +    uint64_t sample_pages_per_gigabytes = config->sample_pages_per_gigabytes;
> > +
> > +    /* Right shift 30 bits to calc block size in GB */
> > +    info->sample_pages_count = (qemu_ram_get_used_length(block) *
> > +                                sample_pages_per_gigabytes) >>
> > +                                DIRTYRATE_PAGE_SHIFT_GB;
> 
> Doing the calculation this way around seems odd. Why was it not:
> 
>     info->sample_pages_count = (qemu_ram_get_used_length(block) >>
>                                 DIRTYRATE_PAGE_SHIFT_GB) *
>                                 sample_pages_per_gigabytes;
> 
> ?

Because that would give 0 for a 0.5GB block

Dave

> > +
> > +    /* Right shift 12 bits to calc page count in 4KB */
> > +    info->ramblock_pages = qemu_ram_get_used_length(block) >>
> > +                           DIRTYRATE_PAGE_SHIFT_KB;
> > +    info->ramblock_addr = qemu_ram_get_host_addr(block);
> > +    strcpy(info->idstr, qemu_ram_get_idstr(block));
> > +}
> > +
> > +static struct RamblockDirtyInfo *
> > +alloc_ramblock_dirty_info(int *block_index,
> > +                          struct RamblockDirtyInfo *block_dinfo)
> > +{
> > +    struct RamblockDirtyInfo *info = NULL;
> > +    int index = *block_index;
> > +
> > +    if (!block_dinfo) {
> > +        index = 0;
> > +        block_dinfo = g_try_new(struct RamblockDirtyInfo, 1);
> > +    } else {
> > +        index++;
> > +        block_dinfo = g_try_realloc(block_dinfo, (index + 1) *
> > +                                    sizeof(struct RamblockDirtyInfo));
> 
> g_try_renew() instead of g_try_realloc()?
> 
> > +    }
> > +    if (!block_dinfo) {
> > +        return NULL;
> > +    }
> > +
> > +    info = &block_dinfo[index];
> > +    *block_index = index;
> > +    memset(info, 0, sizeof(struct RamblockDirtyInfo));
> > +
> > +    return block_dinfo;
> > +}
> > +
> > +static int record_ramblock_hash_info(struct RamblockDirtyInfo **block_dinfo,
> > +                                     struct DirtyRateConfig config,
> > +                                     int *block_index)
> > +{
> > +    struct RamblockDirtyInfo *info = NULL;
> > +    struct RamblockDirtyInfo *dinfo = NULL;
> > +    RAMBlock *block = NULL;
> 
> There's no need to initialise "info" or "block".
> 
> The declaration of "info" could be pushed into the loop.
> 
> > +    int index = 0;
> > +
> > +    RAMBLOCK_FOREACH_MIGRATABLE(block) {
> > +        dinfo = alloc_ramblock_dirty_info(&index, dinfo);
> > +        if (dinfo == NULL) {
> > +            return -1;
> > +        }
> > +        info = &dinfo[index];
> > +        get_ramblock_dirty_info(block, info, &config);
> > +        if (save_ramblock_hash(info) < 0) {
> > +            *block_dinfo = dinfo;
> > +            *block_index = index;
> > +            return -1;
> > +        }
> > +    }
> > +
> > +    *block_dinfo = dinfo;
> > +    *block_index = index;
> > +
> > +    return 0;
> > +}
> > +
> >  static void calculate_dirtyrate(struct DirtyRateConfig config)
> >  {
> >      /* todo */
> > diff --git a/migration/dirtyrate.h b/migration/dirtyrate.h
> > index 9db269d..5050add 100644
> > --- a/migration/dirtyrate.h
> > +++ b/migration/dirtyrate.h
> > @@ -24,6 +24,21 @@
> >   */
> >  #define RAMBLOCK_INFO_MAX_LEN                     256
> >  
> > +/*
> > + * Sample page size 4K as default.
> > + */
> > +#define DIRTYRATE_SAMPLE_PAGE_SIZE                4096
> > +
> > +/*
> > + * Sample page size 4K shift
> > + */
> > +#define DIRTYRATE_PAGE_SHIFT_KB                   12
> > +
> > +/*
> > + * Sample page size 1G shift
> > + */
> > +#define DIRTYRATE_PAGE_SHIFT_GB                   30
> > +
> >  /* Take 1s as default for calculation duration */
> >  #define DEFAULT_FETCH_DIRTYRATE_TIME_SEC          1
> >  
> > -- 
> > 1.8.3.1
> 
> dme.
> -- 
> You bring light in.
>
David Edmondson Aug. 26, 2020, 12:33 p.m. UTC | #3
On Wednesday, 2020-08-26 at 13:30:16 +01, Dr. David Alan Gilbert wrote:

> * David Edmondson (dme@dme.org) wrote:
>> On Monday, 2020-08-24 at 17:14:34 +08, Chuan Zheng wrote:
>> 
>> > Record hash results for each sampled page, crc32 is taken to calculate
>> > hash results for each sampled 4K-page.
>> >
>> > Signed-off-by: Chuan Zheng <zhengchuan@huawei.com>
>> > Signed-off-by: YanYing Zhuang <ann.zhuangyanying@huawei.com>
>> > ---
>> >  migration/dirtyrate.c | 136 ++++++++++++++++++++++++++++++++++++++++++++++++++
>> >  migration/dirtyrate.h |  15 ++++++
>> >  2 files changed, 151 insertions(+)
>> >
>> > diff --git a/migration/dirtyrate.c b/migration/dirtyrate.c
>> > index f6a94d8..66de426 100644
>> > --- a/migration/dirtyrate.c
>> > +++ b/migration/dirtyrate.c
>> > @@ -10,6 +10,7 @@
>> >   * See the COPYING file in the top-level directory.
>> >   */
>> >  
>> > +#include <zlib.h>
>> >  #include "qemu/osdep.h"
>> >  #include "qapi/error.h"
>> >  #include "crypto/hash.h"
>> > @@ -66,6 +67,141 @@ static void update_dirtyrate(uint64_t msec)
>> >      DirtyStat.dirty_rate = dirtyrate;
>> >  }
>> >  
>> > +/*
>> > + * get hash result for the sampled memory with length of 4K byte in ramblock,
>> > + * which starts from ramblock base address.
>> > + */
>> > +static uint32_t get_ramblock_vfn_hash(struct RamblockDirtyInfo *info,
>> > +                                      uint64_t vfn)
>> > +{
>> > +    struct iovec iov_array;
>> 
>> There's no need for an iovec now that crc32() is being used.
>> 
>> > +    uint32_t crc;
>> > +
>> > +    iov_array.iov_base = info->ramblock_addr +
>> > +                         vfn * DIRTYRATE_SAMPLE_PAGE_SIZE;
>> > +    iov_array.iov_len = DIRTYRATE_SAMPLE_PAGE_SIZE;
>> > +
>> > +    crc = crc32(0, iov_array.iov_base, iov_array.iov_len);
>> > +
>> > +    return crc;
>> > +}
>> > +
>> > +static int save_ramblock_hash(struct RamblockDirtyInfo *info)
>> > +{
>> > +    unsigned int sample_pages_count;
>> > +    int i;
>> > +    int ret = -1;
>> 
>> There's no need to initialise "ret".
>> 
>> > +    GRand *rand = g_rand_new();
>> 
>> Calling g_rand_new() when the result may not be used (because of the
>> various conditions immediately below) seems as though it might waste
>> entropy. Could this be pushed down just above the loop? It would even
>> get rid of the gotos ;-)
>> 
>> > +    sample_pages_count = info->sample_pages_count;
>> > +
>> > +    /* ramblock size less than one page, return success to skip this ramblock */
>> > +    if (unlikely(info->ramblock_pages == 0 || sample_pages_count == 0)) {
>> > +        ret = 0;
>> > +        goto out;
>> > +    }
>> > +
>> > +    info->hash_result = g_try_malloc0_n(sample_pages_count,
>> > +                                        sizeof(uint32_t));
>> > +    if (!info->hash_result) {
>> > +        ret = -1;
>> > +        goto out;
>> > +    }
>> > +
>> > +    info->sample_page_vfn = g_try_malloc0_n(sample_pages_count,
>> > +                                            sizeof(uint64_t));
>> > +    if (!info->sample_page_vfn) {
>> > +        g_free(info->hash_result);
>> > +        ret = -1;
>> > +        goto out;
>> > +    }
>> > +
>> > +    for (i = 0; i < sample_pages_count; i++) {
>> > +        info->sample_page_vfn[i] = g_rand_int_range(rand, 0,
>> > +                                                    info->ramblock_pages - 1);
>> > +        info->hash_result[i] = get_ramblock_vfn_hash(info,
>> > +                                                     info->sample_page_vfn[i]);
>> > +    }
>> > +    ret = 0;
>> > +
>> > +out:
>> > +    g_rand_free(rand);
>> > +    return ret;
>> > +}
>> > +
>> > +static void get_ramblock_dirty_info(RAMBlock *block,
>> > +                                    struct RamblockDirtyInfo *info,
>> > +                                    struct DirtyRateConfig *config)
>> > +{
>> > +    uint64_t sample_pages_per_gigabytes = config->sample_pages_per_gigabytes;
>> > +
>> > +    /* Right shift 30 bits to calc block size in GB */
>> > +    info->sample_pages_count = (qemu_ram_get_used_length(block) *
>> > +                                sample_pages_per_gigabytes) >>
>> > +                                DIRTYRATE_PAGE_SHIFT_GB;
>> 
>> Doing the calculation this way around seems odd. Why was it not:
>> 
>>     info->sample_pages_count = (qemu_ram_get_used_length(block) >>
>>                                 DIRTYRATE_PAGE_SHIFT_GB) *
>>                                 sample_pages_per_gigabytes;
>> 
>> ?
>
> Because that would give 0 for a 0.5GB block

Ouch, obviously :-) Thanks.

dme.
Zheng Chuan Aug. 27, 2020, 6:28 a.m. UTC | #4
On 2020/8/26 20:30, Dr. David Alan Gilbert wrote:
> * David Edmondson (dme@dme.org) wrote:
>> On Monday, 2020-08-24 at 17:14:34 +08, Chuan Zheng wrote:
>>
>>> Record hash results for each sampled page, crc32 is taken to calculate
>>> hash results for each sampled 4K-page.
>>>
>>> Signed-off-by: Chuan Zheng <zhengchuan@huawei.com>
>>> Signed-off-by: YanYing Zhuang <ann.zhuangyanying@huawei.com>
>>> ---
>>>  migration/dirtyrate.c | 136 ++++++++++++++++++++++++++++++++++++++++++++++++++
>>>  migration/dirtyrate.h |  15 ++++++
>>>  2 files changed, 151 insertions(+)
>>>
>>> diff --git a/migration/dirtyrate.c b/migration/dirtyrate.c
>>> index f6a94d8..66de426 100644
>>> --- a/migration/dirtyrate.c
>>> +++ b/migration/dirtyrate.c
>>> @@ -10,6 +10,7 @@
>>>   * See the COPYING file in the top-level directory.
>>>   */
>>>  
>>> +#include <zlib.h>
>>>  #include "qemu/osdep.h"
>>>  #include "qapi/error.h"
>>>  #include "crypto/hash.h"
>>> @@ -66,6 +67,141 @@ static void update_dirtyrate(uint64_t msec)
>>>      DirtyStat.dirty_rate = dirtyrate;
>>>  }
>>>  
>>> +/*
>>> + * get hash result for the sampled memory with length of 4K byte in ramblock,
>>> + * which starts from ramblock base address.
>>> + */
>>> +static uint32_t get_ramblock_vfn_hash(struct RamblockDirtyInfo *info,
>>> +                                      uint64_t vfn)
>>> +{
>>> +    struct iovec iov_array;
>>
>> There's no need for an iovec now that crc32() is being used.
>>
OK, will take two variables instead to represent base address and length.

>>> +    uint32_t crc;
>>> +
>>> +    iov_array.iov_base = info->ramblock_addr +
>>> +                         vfn * DIRTYRATE_SAMPLE_PAGE_SIZE;
>>> +    iov_array.iov_len = DIRTYRATE_SAMPLE_PAGE_SIZE;
>>> +
>>> +    crc = crc32(0, iov_array.iov_base, iov_array.iov_len);
>>> +
>>> +    return crc;
>>> +}
>>> +
>>> +static int save_ramblock_hash(struct RamblockDirtyInfo *info)
>>> +{
>>> +    unsigned int sample_pages_count;
>>> +    int i;
>>> +    int ret = -1;
>>
>> There's no need to initialise "ret".
>>
>>> +    GRand *rand = g_rand_new();
>>
>> Calling g_rand_new() when the result may not be used (because of the
>> various conditions immediately below) seems as though it might waste
>> entropy. Could this be pushed down just above the loop? It would even
>> get rid of the gotos ;-)
>>
OK, i'll try it.

>>> +    sample_pages_count = info->sample_pages_count;
>>> +
>>> +    /* ramblock size less than one page, return success to skip this ramblock */
>>> +    if (unlikely(info->ramblock_pages == 0 || sample_pages_count == 0)) {
>>> +        ret = 0;
>>> +        goto out;
>>> +    }
>>> +
>>> +    info->hash_result = g_try_malloc0_n(sample_pages_count,
>>> +                                        sizeof(uint32_t));
>>> +    if (!info->hash_result) {
>>> +        ret = -1;
>>> +        goto out;
>>> +    }
>>> +
>>> +    info->sample_page_vfn = g_try_malloc0_n(sample_pages_count,
>>> +                                            sizeof(uint64_t));
>>> +    if (!info->sample_page_vfn) {
>>> +        g_free(info->hash_result);
>>> +        ret = -1;
>>> +        goto out;
>>> +    }
>>> +
>>> +    for (i = 0; i < sample_pages_count; i++) {
>>> +        info->sample_page_vfn[i] = g_rand_int_range(rand, 0,
>>> +                                                    info->ramblock_pages - 1);
>>> +        info->hash_result[i] = get_ramblock_vfn_hash(info,
>>> +                                                     info->sample_page_vfn[i]);
>>> +    }
>>> +    ret = 0;
>>> +
>>> +out:
>>> +    g_rand_free(rand);
>>> +    return ret;
>>> +}
>>> +
>>> +static void get_ramblock_dirty_info(RAMBlock *block,
>>> +                                    struct RamblockDirtyInfo *info,
>>> +                                    struct DirtyRateConfig *config)
>>> +{
>>> +    uint64_t sample_pages_per_gigabytes = config->sample_pages_per_gigabytes;
>>> +
>>> +    /* Right shift 30 bits to calc block size in GB */
>>> +    info->sample_pages_count = (qemu_ram_get_used_length(block) *
>>> +                                sample_pages_per_gigabytes) >>
>>> +                                DIRTYRATE_PAGE_SHIFT_GB;
>>
>> Doing the calculation this way around seems odd. Why was it not:
>>
>>     info->sample_pages_count = (qemu_ram_get_used_length(block) >>
>>                                 DIRTYRATE_PAGE_SHIFT_GB) *
>>                                 sample_pages_per_gigabytes;
>>
>> ?
> 
> Because that would give 0 for a 0.5GB block
> 
> Dave
> 
>>> +
>>> +    /* Right shift 12 bits to calc page count in 4KB */
>>> +    info->ramblock_pages = qemu_ram_get_used_length(block) >>
>>> +                           DIRTYRATE_PAGE_SHIFT_KB;
>>> +    info->ramblock_addr = qemu_ram_get_host_addr(block);
>>> +    strcpy(info->idstr, qemu_ram_get_idstr(block));
>>> +}
>>> +
>>> +static struct RamblockDirtyInfo *
>>> +alloc_ramblock_dirty_info(int *block_index,
>>> +                          struct RamblockDirtyInfo *block_dinfo)
>>> +{
>>> +    struct RamblockDirtyInfo *info = NULL;
>>> +    int index = *block_index;
>>> +
>>> +    if (!block_dinfo) {
>>> +        index = 0;
>>> +        block_dinfo = g_try_new(struct RamblockDirtyInfo, 1);
>>> +    } else {
>>> +        index++;
>>> +        block_dinfo = g_try_realloc(block_dinfo, (index + 1) *
>>> +                                    sizeof(struct RamblockDirtyInfo));
>>
>> g_try_renew() instead of g_try_realloc()?
>>
Hi,
I am not sure that because there only one place in qemu to use g_try_renew.
Could you tell me why, because i think g_try_realloc will also return NULL when error happen:)

>>> +    }
>>> +    if (!block_dinfo) {
>>> +        return NULL;
>>> +    }
>>> +
>>> +    info = &block_dinfo[index];
>>> +    *block_index = index;
>>> +    memset(info, 0, sizeof(struct RamblockDirtyInfo));
>>> +
>>> +    return block_dinfo;
>>> +}
>>> +
>>> +static int record_ramblock_hash_info(struct RamblockDirtyInfo **block_dinfo,
>>> +                                     struct DirtyRateConfig config,
>>> +                                     int *block_index)
>>> +{
>>> +    struct RamblockDirtyInfo *info = NULL;
>>> +    struct RamblockDirtyInfo *dinfo = NULL;
>>> +    RAMBlock *block = NULL;
>>
>> There's no need to initialise "info" or "block".
>>
OK, will remove unused initialization in V6.
>> The declaration of "info" could be pushed into the loop.
>>
>>> +    int index = 0;
>>> +
>>> +    RAMBLOCK_FOREACH_MIGRATABLE(block) {
>>> +        dinfo = alloc_ramblock_dirty_info(&index, dinfo);
>>> +        if (dinfo == NULL) {
>>> +            return -1;
>>> +        }
>>> +        info = &dinfo[index];
>>> +        get_ramblock_dirty_info(block, info, &config);
>>> +        if (save_ramblock_hash(info) < 0) {
>>> +            *block_dinfo = dinfo;
>>> +            *block_index = index;
>>> +            return -1;
>>> +        }
>>> +    }
>>> +
>>> +    *block_dinfo = dinfo;
>>> +    *block_index = index;
>>> +
>>> +    return 0;
>>> +}
>>> +
>>>  static void calculate_dirtyrate(struct DirtyRateConfig config)
>>>  {
>>>      /* todo */
>>> diff --git a/migration/dirtyrate.h b/migration/dirtyrate.h
>>> index 9db269d..5050add 100644
>>> --- a/migration/dirtyrate.h
>>> +++ b/migration/dirtyrate.h
>>> @@ -24,6 +24,21 @@
>>>   */
>>>  #define RAMBLOCK_INFO_MAX_LEN                     256
>>>  
>>> +/*
>>> + * Sample page size 4K as default.
>>> + */
>>> +#define DIRTYRATE_SAMPLE_PAGE_SIZE                4096
>>> +
>>> +/*
>>> + * Sample page size 4K shift
>>> + */
>>> +#define DIRTYRATE_PAGE_SHIFT_KB                   12
>>> +
>>> +/*
>>> + * Sample page size 1G shift
>>> + */
>>> +#define DIRTYRATE_PAGE_SHIFT_GB                   30
>>> +
>>>  /* Take 1s as default for calculation duration */
>>>  #define DEFAULT_FETCH_DIRTYRATE_TIME_SEC          1
>>>  
>>> -- 
>>> 1.8.3.1
>>
>> dme.
>> -- 
>> You bring light in.
>>
David Edmondson Aug. 27, 2020, 7:11 a.m. UTC | #5
On Thursday, 2020-08-27 at 14:28:03 +08, Zheng Chuan wrote:

>>>> +static struct RamblockDirtyInfo *
>>>> +alloc_ramblock_dirty_info(int *block_index,
>>>> +                          struct RamblockDirtyInfo *block_dinfo)
>>>> +{
>>>> +    struct RamblockDirtyInfo *info = NULL;
>>>> +    int index = *block_index;
>>>> +
>>>> +    if (!block_dinfo) {
>>>> +        index = 0;
>>>> +        block_dinfo = g_try_new(struct RamblockDirtyInfo, 1);
>>>> +    } else {
>>>> +        index++;
>>>> +        block_dinfo = g_try_realloc(block_dinfo, (index + 1) *
>>>> +                                    sizeof(struct RamblockDirtyInfo));
>>>
>>> g_try_renew() instead of g_try_realloc()?
>>>
> Hi,
> I am not sure that because there only one place in qemu to use g_try_renew.
> Could you tell me why, because i think g_try_realloc will also return NULL when error happen:)

Only suggested because it would make the two branches of the code more
similar.

dme.
diff mbox series

Patch

diff --git a/migration/dirtyrate.c b/migration/dirtyrate.c
index f6a94d8..66de426 100644
--- a/migration/dirtyrate.c
+++ b/migration/dirtyrate.c
@@ -10,6 +10,7 @@ 
  * See the COPYING file in the top-level directory.
  */
 
+#include <zlib.h>
 #include "qemu/osdep.h"
 #include "qapi/error.h"
 #include "crypto/hash.h"
@@ -66,6 +67,141 @@  static void update_dirtyrate(uint64_t msec)
     DirtyStat.dirty_rate = dirtyrate;
 }
 
+/*
+ * get hash result for the sampled memory with length of 4K byte in ramblock,
+ * which starts from ramblock base address.
+ */
+static uint32_t get_ramblock_vfn_hash(struct RamblockDirtyInfo *info,
+                                      uint64_t vfn)
+{
+    struct iovec iov_array;
+    uint32_t crc;
+
+    iov_array.iov_base = info->ramblock_addr +
+                         vfn * DIRTYRATE_SAMPLE_PAGE_SIZE;
+    iov_array.iov_len = DIRTYRATE_SAMPLE_PAGE_SIZE;
+
+    crc = crc32(0, iov_array.iov_base, iov_array.iov_len);
+
+    return crc;
+}
+
+static int save_ramblock_hash(struct RamblockDirtyInfo *info)
+{
+    unsigned int sample_pages_count;
+    int i;
+    int ret = -1;
+    GRand *rand = g_rand_new();
+
+    sample_pages_count = info->sample_pages_count;
+
+    /* ramblock size less than one page, return success to skip this ramblock */
+    if (unlikely(info->ramblock_pages == 0 || sample_pages_count == 0)) {
+        ret = 0;
+        goto out;
+    }
+
+    info->hash_result = g_try_malloc0_n(sample_pages_count,
+                                        sizeof(uint32_t));
+    if (!info->hash_result) {
+        ret = -1;
+        goto out;
+    }
+
+    info->sample_page_vfn = g_try_malloc0_n(sample_pages_count,
+                                            sizeof(uint64_t));
+    if (!info->sample_page_vfn) {
+        g_free(info->hash_result);
+        ret = -1;
+        goto out;
+    }
+
+    for (i = 0; i < sample_pages_count; i++) {
+        info->sample_page_vfn[i] = g_rand_int_range(rand, 0,
+                                                    info->ramblock_pages - 1);
+        info->hash_result[i] = get_ramblock_vfn_hash(info,
+                                                     info->sample_page_vfn[i]);
+    }
+    ret = 0;
+
+out:
+    g_rand_free(rand);
+    return ret;
+}
+
+static void get_ramblock_dirty_info(RAMBlock *block,
+                                    struct RamblockDirtyInfo *info,
+                                    struct DirtyRateConfig *config)
+{
+    uint64_t sample_pages_per_gigabytes = config->sample_pages_per_gigabytes;
+
+    /* Right shift 30 bits to calc block size in GB */
+    info->sample_pages_count = (qemu_ram_get_used_length(block) *
+                                sample_pages_per_gigabytes) >>
+                                DIRTYRATE_PAGE_SHIFT_GB;
+
+    /* Right shift 12 bits to calc page count in 4KB */
+    info->ramblock_pages = qemu_ram_get_used_length(block) >>
+                           DIRTYRATE_PAGE_SHIFT_KB;
+    info->ramblock_addr = qemu_ram_get_host_addr(block);
+    strcpy(info->idstr, qemu_ram_get_idstr(block));
+}
+
+static struct RamblockDirtyInfo *
+alloc_ramblock_dirty_info(int *block_index,
+                          struct RamblockDirtyInfo *block_dinfo)
+{
+    struct RamblockDirtyInfo *info = NULL;
+    int index = *block_index;
+
+    if (!block_dinfo) {
+        index = 0;
+        block_dinfo = g_try_new(struct RamblockDirtyInfo, 1);
+    } else {
+        index++;
+        block_dinfo = g_try_realloc(block_dinfo, (index + 1) *
+                                    sizeof(struct RamblockDirtyInfo));
+    }
+    if (!block_dinfo) {
+        return NULL;
+    }
+
+    info = &block_dinfo[index];
+    *block_index = index;
+    memset(info, 0, sizeof(struct RamblockDirtyInfo));
+
+    return block_dinfo;
+}
+
+static int record_ramblock_hash_info(struct RamblockDirtyInfo **block_dinfo,
+                                     struct DirtyRateConfig config,
+                                     int *block_index)
+{
+    struct RamblockDirtyInfo *info = NULL;
+    struct RamblockDirtyInfo *dinfo = NULL;
+    RAMBlock *block = NULL;
+    int index = 0;
+
+    RAMBLOCK_FOREACH_MIGRATABLE(block) {
+        dinfo = alloc_ramblock_dirty_info(&index, dinfo);
+        if (dinfo == NULL) {
+            return -1;
+        }
+        info = &dinfo[index];
+        get_ramblock_dirty_info(block, info, &config);
+        if (save_ramblock_hash(info) < 0) {
+            *block_dinfo = dinfo;
+            *block_index = index;
+            return -1;
+        }
+    }
+
+    *block_dinfo = dinfo;
+    *block_index = index;
+
+    return 0;
+}
+
 static void calculate_dirtyrate(struct DirtyRateConfig config)
 {
     /* todo */
diff --git a/migration/dirtyrate.h b/migration/dirtyrate.h
index 9db269d..5050add 100644
--- a/migration/dirtyrate.h
+++ b/migration/dirtyrate.h
@@ -24,6 +24,21 @@ 
  */
 #define RAMBLOCK_INFO_MAX_LEN                     256
 
+/*
+ * Sample page size 4K as default.
+ */
+#define DIRTYRATE_SAMPLE_PAGE_SIZE                4096
+
+/*
+ * Sample page size 4K shift
+ */
+#define DIRTYRATE_PAGE_SHIFT_KB                   12
+
+/*
+ * Sample page size 1G shift
+ */
+#define DIRTYRATE_PAGE_SHIFT_GB                   30
+
 /* Take 1s as default for calculation duration */
 #define DEFAULT_FETCH_DIRTYRATE_TIME_SEC          1