Re: [PATCH v2 04/19] block/nvme: Trace controller capabilities

[Stefan Hajnoczi <stefanha@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2186 bytes --]

On Mon, Oct 26, 2020 at 11:54:49AM +0100, Philippe Mathieu-Daudé wrote:
> Controllers have different capabilities and report them in the
> CAP register. We are particularly interested by the page size
> limits.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
> ---
>  block/nvme.c       | 10 ++++++++++
>  block/trace-events |  1 +
>  2 files changed, 11 insertions(+)
> 
> diff --git a/block/nvme.c b/block/nvme.c
> index 5abd7257cac..3b6d3972ec2 100644
> --- a/block/nvme.c
> +++ b/block/nvme.c
> @@ -720,6 +720,16 @@ static int nvme_init(BlockDriverState *bs, const char *device, int namespace,
>       * Initialization". */
>  
>      cap = le64_to_cpu(regs->cap);
> +    trace_nvme_controller_capability("Maximum Queue Entries Supported",
> +                                     NVME_CAP_MQES(cap));
> +    trace_nvme_controller_capability("Contiguous Queues Required",
> +                                     NVME_CAP_CQR(cap));
> +    trace_nvme_controller_capability("Subsystem  Reset Supported",
> +                                     NVME_CAP_NSSRS(cap));
> +    trace_nvme_controller_capability("Memory Page Size Minimum",
> +                                     NVME_CAP_MPSMIN(cap));
> +    trace_nvme_controller_capability("Memory Page Size Maximum",
> +                                     NVME_CAP_MPSMAX(cap));

This works well for printf-style tracing but it can be tedious to
strcmp() or parse strings from SystemTap and other tracing scripts.

Another way of expressing these trace events is:

  trace_nvme_controller_capabilities(NVME_CAP_MQES(cap),
                                     NVME_CAP_CQR(cap),
				     ...);

Or:

  trace_nvme_controller_capability_mqes(NVME_CAP_MQES(cap));
  trace_nvme_controller_capability_cqr(NVME_CAP_CQR(cap));

One view is that tracing is like printf (the style used in this patch).
Another view is that tracing produces records that scripts can process
(the style I showed). Trace events defined in one style may be hard to
use in the other style (i.e. less human-readable vs harder to process in
a script).

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]