On 10/31/22 7:41 AM, Dylan Yudaken wrote:
> Requests can be in flight on the io-wq, and can be long lived (for example
> a double read will get onto the io-wq). So make sure to retarget the rsrc
> nodes on those requests.
>
> Signed-off-by: Dylan Yudaken <dylany@xxxxxxxx>
> ---
> io_uring/rsrc.c | 46 ++++++++++++++++++++++++++++++++++++++++++----
> 1 file changed, 42 insertions(+), 4 deletions(-)
>
> diff --git a/io_uring/rsrc.c b/io_uring/rsrc.c
> index 106210e0d5d5..8d0d40713a63 100644
> --- a/io_uring/rsrc.c
> +++ b/io_uring/rsrc.c
> @@ -16,6 +16,7 @@
> #include "openclose.h"
> #include "rsrc.h"
> #include "opdef.h"
> +#include "tctx.h"
>
> struct io_rsrc_update {
> struct file *file;
> @@ -24,6 +25,11 @@ struct io_rsrc_update {
> u32 offset;
> };
>
> +struct io_retarget_data {
> + struct io_ring_ctx *ctx;
> + unsigned int refs;
> +};
Do we really need this struct? As far as I can tell, you pass in ctx
only and back refs. It's passed in the callbacks, but they only care
about ctx. If io_rsrc_retarget_wq() returned back the refs rather than
use data->refs, then we could just pass in the ctx?
Or you could at least keep it local to io_rsrc_retarget_wq() and
io_retarget_rsrc_wq_cb().
Not a big deal, just always nice to keep the scope of struct as small as
can be (or get rid of them).
--
Jens Axboe
