On Thu, Sep 25, 2025 at 05:02:25PM +0200, Johannes Thumshirn wrote:
> Add definitions for the extended version of the blktrace protocol using a
> wider action type to be able to record new actions in the kernel.
> 
> Signed-off-by: Johannes Thumshirn <[email protected]>
> ---
>  include/uapi/linux/blktrace_api.h | 16 ++++++++++++++++
>  1 file changed, 16 insertions(+)
> 
> diff --git a/include/uapi/linux/blktrace_api.h 
> b/include/uapi/linux/blktrace_api.h
> index a3b1f35ac026..d58ef484de49 100644
> --- a/include/uapi/linux/blktrace_api.h
> +++ b/include/uapi/linux/blktrace_api.h
> @@ -94,6 +94,7 @@ enum blktrace_notify {
>  
>  #define BLK_IO_TRACE_MAGIC   0x65617400
>  #define BLK_IO_TRACE_VERSION 0x07
> +#define BLK_IO_TRACE2_VERSION        0x08
>  
>  /*
>   * The trace itself
> @@ -113,6 +114,21 @@ struct blk_io_trace {
>       /* cgroup id will be stored here if exists */
>  };
>  
> +struct blk_io_trace2 {
> +     __u32 magic;            /* MAGIC << 8 | BLK_IO_TRACE2_VERSION */
> +     __u32 sequence;         /* event number */
> +     __u64 time;             /* in nanoseconds */
> +     __u64 sector;           /* disk offset */
> +     __u32 bytes;            /* transfer length */
> +     __u32 pid;              /* who did it */
> +     __u64 action;           /* what happened */
> +     __u32 device;           /* device number */
> +     __u32 cpu;              /* on what cpu did it happen */
> +     __u16 error;            /* completion error */
> +     __u16 pdu_len;          /* length of data after this trace */
> +     __u8 pad[8];

This will cause mismatching sizes between x86_32 and other
architectures because the size is not 8-byte aligned.  You'll need
to add or remove 4 bytes of padding to ensure that.

I really wish we could have good helpers to check for that.


Reply via email to