Re: [Qemu-devel] [v8 Patch 6/6]Qemu: raw posix implementation of reopen functions

[Kevin Wolf <kwolf@redhat.com>]

Am 30.10.2011 11:35, schrieb Supriya Kannery:
> raw-posix driver changes for bdrv_reopen_xx functions to
> safely reopen image files. Reopening of image files while 
> changing hostcache dynamically, is handled here.
> 
> Signed-off-by: Supriya Kannery <supriyak@linux.vnet.ibm.com>
> 
> Index: qemu/block/raw.c
> ===================================================================
> --- qemu.orig/block/raw.c
> +++ qemu/block/raw.c
> @@ -9,6 +9,24 @@ static int raw_open(BlockDriverState *bs
>      return 0;
>  }
>  
> +static int raw_reopen_prepare(BlockDriverState *bs, BDRVReopenState **prs,
> +                              int flags)
> +{
> +    return bdrv_reopen_prepare(bs->file, prs, flags);
> +}
> +
> +static void raw_reopen_commit(BlockDriverState *bs, BDRVReopenState *rs,
> +                              int flags)
> +{
> +    bdrv_reopen_commit(bs->file, rs, flags);
> +    bs->open_flags = bs->file->open_flags;

I think it should be bs->open_flags = flags, even if the underlying
driver masks something away (which it shouldn't do in the first place,
but using flags here is clearer).

Also I'm wondering if updating bs->open_flags isn't common to all
formats, so could we move it into the block.c function?

> +}
> +
> +static void raw_reopen_abort(BlockDriverState *bs, BDRVReopenState *rs)
> +{
> +    bdrv_reopen_abort(bs->file, rs);
> +}
> +
>  static int coroutine_fn raw_co_readv(BlockDriverState *bs, int64_t sector_num,
>                                       int nb_sectors, QEMUIOVector *qiov)
>  {
> @@ -107,7 +125,10 @@ static BlockDriver bdrv_raw = {
>  
>      /* It's really 0, but we need to make g_malloc() happy */
>      .instance_size      = 1,
> -
> +    .bdrv_reopen_prepare
> +                        = raw_reopen_prepare,
> +    .bdrv_reopen_commit = raw_reopen_commit,
> +    .bdrv_reopen_abort  = raw_reopen_abort,
>      .bdrv_open          = raw_open,
>      .bdrv_close         = raw_close,

Mostly a matter of taste, but I would prefer open/close to stay first
and having bdrv_reopen_* only after them.

>  
> Index: qemu/block/raw-posix.c
> ===================================================================
> --- qemu.orig/block/raw-posix.c
> +++ qemu/block/raw-posix.c
> @@ -279,6 +279,60 @@ static int raw_open(BlockDriverState *bs
>      return raw_open_common(bs, filename, flags, 0);
>  }
>  
> +static int raw_reopen_prepare(BlockDriverState *bs, BDRVReopenState **prs,
> +                              int flags)
> +{
> +    BDRVReopenState *raw_rs = g_malloc0(sizeof(BDRVReopenState));
> +    BDRVRawState *s = bs->opaque;
> +
> +    raw_rs->bs = bs;
> +    raw_rs->reopen_flags = s->open_flags;
> +    raw_rs->reopen_fd = -1;
> +    *prs = raw_rs;
> +    int ret = 0;
> +
> +    /* If only O_DIRECT to be toggled, use fcntl */
> +    if (!((bs->open_flags & ~BDRV_O_NOCACHE) ^
> +            (flags & ~BDRV_O_NOCACHE))) {

Wouldn't it be more readable like this?

/* Use fcntl if all affected flags can be changes this way */
fcntl_flags = BDRV_O_NOCACHE;
if ((bs->open_flags & ~fcntl_flags) == (flags & ~fcntl_flags)) {
    ...


> +        raw_rs->reopen_fd = dup(s->fd);
> +        if (raw_rs->reopen_fd <= 0) {
> +            return -1;
> +        }
> +        if ((flags & BDRV_O_NOCACHE)) {
> +            raw_rs->reopen_flags |= O_DIRECT;
> +        } else {
> +            raw_rs->reopen_flags &= ~O_DIRECT;
> +        }
> +        ret = fcntl_setfl(raw_rs->reopen_fd, raw_rs->reopen_flags);
> +    }

I think there should be an else branch that returns an error. Currently
requests involving anything but BDRV_O_NOCACHE are completely ignored,
but still success is returned.

> +
> +    /* TBD: Handle O_DSYNC and other flags */
> +
> +    return ret;
> +}

Kevin