Re: [PATCH v5 30/38] SQUASHME: initialize data->task on the non-rpc io done success paths

[Boaz Harrosh <bharrosh@panasas.com>]

On 05/23/2011 03:01 AM, Benny Halevy wrote:
> Signed-off-by: Benny Halevy <bhalevy@panasas.com>
> ---
>  fs/nfs/pnfs.c |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)
> 
> diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
> index 0f59802..d39fcca 100644
> --- a/fs/nfs/pnfs.c
> +++ b/fs/nfs/pnfs.c
> @@ -1064,6 +1064,7 @@ pnfs_ld_write_done(struct nfs_write_data *data)
>  
>  	if (!data->pnfs_error) {
>  		pnfs_set_layoutcommit(data);
> +		memset(&data->task, 0, sizeof(data->task));

What?
We used this data task to come here. See: objlayout.c::objlayout_read/write_done()
why do you think it is invalid?

>  		data->mds_ops->rpc_call_done(&data->task, data);
>  		data->mds_ops->rpc_release(data);
>  		return 0;
> @@ -1113,6 +1114,7 @@ pnfs_ld_read_done(struct nfs_read_data *data)
>  
>  	if (!data->pnfs_error) {
>  		__nfs4_read_done_cb(data);
> +		memset(&data->task, 0, sizeof(data->task));

Same here this is called on the task thread

>  		data->mds_ops->rpc_call_done(&data->task, data);
>  		data->mds_ops->rpc_release(data);
>  		return 0;

Boaz