Re: [Qemu-devel] [PATCH 18/24] block: Initial implementation of -blockdev

[Kevin Wolf]

Am 27.02.2017 um 12:20 hat Markus Armbruster geschrieben:
> The new command line option -blockdev works like QMP command
> blockdev-add.
> 
> The option argument may be given in JSON syntax, exactly as in QMP.
> Example usage:
> 
>     -blockdev '{"node-name": "foo", "driver": "raw", "file": {"driver": 
> "file", "filename": "foo.img"} }'
> 
> The JSON argument doesn't exactly blend into the existing option
> syntax, so the traditional KEY=VALUE,... syntax is also supported,
> using dotted keys to do the nesting:
> 
>     -blockdev node-name=foo,driver=raw,file.driver=file,file.filename=foo.img
> 
> This does not yet support lists or downstream extensions, i.e. keys
> with __RFQDN_ prefix, but the next few patches will take care of that.
> 
> Note that calling qmp_blockdev_add() (say via qmp_marshal_block_add())
> right away would crash.  We need to stash the configuration for later
> instead.  This is crudely done, and bypasses QemuOpts, even though
> storing configuration is what QemuOpts is for.  Need to revamp option
> infrastructure to support QAPI types like BlockdevOptions.
> 
> Signed-off-by: Markus Armbruster <address@hidden>

Apparently I forgot to actually reply here after going back to the
previous patch...

> @@ -532,6 +532,13 @@ Use @var{file} as CD-ROM image (you cannot use 
> @option{-hdc} and
>  using @file{/dev/cdrom} as filename (@pxref{host_drives}).
>  ETEXI
>  
> +DEF("blockdev", HAS_ARG, QEMU_OPTION_blockdev,
> +    "-blockdev driver[,node-name=N][,discard=ignore|unmap]\n"

I was going to suggest the same thing as Eric here, simply 'driver'
doesn't look right.

> +    "          [,cache.direct=on|off][,cache.no-flush=on|off]\n"
> +    "          [,read-only=on|off][,detect-zeroes=on|off|unmap]\n"
> +    "          [,driver specific parameters...]\n"
> +    "                configure a block backend\n", QEMU_ARCH_ALL)
> +
>  DEF("drive", HAS_ARG, QEMU_OPTION_drive,
>      "-drive [file=file][,if=type][,bus=n][,unit=m][,media=d][,index=i]\n"
>      "       [,cyls=c,heads=h,secs=s[,trans=t]][,snapshot=on|off]\n"

> @@ -4454,6 +4483,16 @@ int main(int argc, char **argv, char **envp)
>          qemu_opts_foreach(qemu_find_opts("drive"), drive_enable_snapshot,
>                            NULL, NULL);
>      }
> +    while (!QSIMPLEQ_EMPTY(&bdo_queue)) {
> +        BlockdevOptions_queue *bdo = QSIMPLEQ_FIRST(&bdo_queue);
> +
> +        QSIMPLEQ_REMOVE_HEAD(&bdo_queue, entry);
> +        loc_push_restore(&bdo->loc);
> +        qmp_blockdev_add(bdo->bdo, &error_fatal);
> +        loc_pop(&bdo->loc);
> +        qapi_free_BlockdevOptions(bdo->bdo);
> +        g_free(bdo);
> +    }
>      if (qemu_opts_foreach(qemu_find_opts("drive"), drive_init_func,
>                            &machine_class->block_default_type, NULL)) {
>          exit(1);

And here I found the placement of the new code odd, it's in the middle
of -drive handling. The current placement means that -drive can
reference -blockdev nodes, but not the other way round. I like this and
wouldn't recommand changing it. So my suggestion for a better placement
would be to move the -blockdev code up a bit.

Kevin