[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v2 5/8] Use writev ops if available
From: |
Orit Wasserman |
Subject: |
Re: [Qemu-devel] [PATCH v2 5/8] Use writev ops if available |
Date: |
Thu, 21 Mar 2013 15:52:34 +0200 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130110 Thunderbird/17.0.2 |
On 03/21/2013 03:44 PM, Paolo Bonzini wrote:
> Il 21/03/2013 14:23, Orit Wasserman ha scritto:
>> Update qemu_fflush and stdio_close to use writev ops if they are available
>>
>> Signed-off-by: Orit Wasserman <address@hidden>
>> ---
>> savevm.c | 30 ++++++++++++++++++++----------
>> 1 file changed, 20 insertions(+), 10 deletions(-)
>>
>> diff --git a/savevm.c b/savevm.c
>> index ab81dd3..fde59d3 100644
>> --- a/savevm.c
>> +++ b/savevm.c
>> @@ -292,7 +292,7 @@ static int stdio_fclose(void *opaque)
>> QEMUFileStdio *s = opaque;
>> int ret = 0;
>>
>> - if (s->file->ops->put_buffer) {
>> + if (s->file->ops->put_buffer || s->file->ops->writev_buffer) {
>> int fd = fileno(s->stdio_file);
>> struct stat st;
>>
>> @@ -515,24 +515,34 @@ static void qemu_file_set_error(QEMUFile *f, int ret)
>> }
>> }
>>
>> -/** Flushes QEMUFile buffer
>> +/**
>> + * Flushes QEMUFile buffer
>> *
>> + * If there is writev_buffer QEMUFileOps it uses it otherwise uses
>> + * put_buffer ops.
>> */
>> static void qemu_fflush(QEMUFile *f)
>> {
>> int ret = 0;
>>
>> - if (!f->ops->put_buffer) {
>> + if (f->ops->writev_buffer) {
>> + if (f->is_write && f->iovcnt > 0) {
>> + ret = f->ops->writev_buffer(f->opaque, f->iov, f->iovcnt);
>
> This needs to update f->pos.
>
>> + }
>> + } else if (f->ops->put_buffer) {
>> + if (f->is_write && f->buf_index > 0) {
>> + ret = f->ops->put_buffer(f->opaque, f->buf, f->pos,
>> f->buf_index);
>> + }
>
> I think this has to go through all the iovecs (i.e. never look at
> f->buf_index, only f->iovcnt). Otherwise RAM migration does not work,
> because the page data is not copied to f->buf. But that's pretty much
> it, the series looks good.
You are right I will fix it.
>
> However, I'd still prefer to have qemu_put_buffer_no_copy coalesce
> adjacent iovecs. Otherwise you might end up with only 3-400 bytes in
> each sendmsg when serializing device data.
Yes this will be helpful for the device state.
Orit
>
> Paolo
>
>> + } else {
>> return;
>> }
>> - if (f->is_write && f->buf_index > 0) {
>> - ret = f->ops->put_buffer(f->opaque, f->buf, f->pos, f->buf_index);
>> - if (ret >= 0) {
>> - f->pos += f->buf_index;
>> - }
>> - f->buf_index = 0;
>> - f->iovcnt = 0;
>> +
>> + if (ret >= 0) {
>> + f->pos += f->buf_index;
>> }
>> + f->buf_index = 0;
>> + f->iovcnt = 0;
>> +
>> if (ret < 0) {
>> qemu_file_set_error(f, ret);
>> }
>>
>
- [Qemu-devel] [PATCH v2 0/8] Migration: Remove copying of guest ram pages, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 2/8] Add socket_writev_buffer function, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 1/8] Add QemuFileWritevBuffer QemuFileOps, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 3/8] Update bytes_xfer in qemu_put_byte, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 6/8] More optimized qemu_put_be64/32/16, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 4/8] Store the data to send also in iovec, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 5/8] Use writev ops if available, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 7/8] Add qemu_put_buffer_no_copy, Orit Wasserman, 2013/03/21
- [Qemu-devel] [PATCH v2 8/8] Use qemu_put_buffer_no_copy for guest memory pages, Orit Wasserman, 2013/03/21