Re: [SCM] systemtap: system-wide probe/trace tool branch, master, updated. release-0.7-276-g707e14f

Re: [SCM] systemtap: system-wide probe/trace tool branch, master, updated. release-0.7-276-g707e14f

[Mark Wielaard]

[Sorry about the resent, apparently the systemtap-cvs list doesn't
accept email. Would it be possible to set the reply-to on those emails
to the main systemtap list in case someone does want to reply to them?]

Hi Zhaolei,

On Mon, 2008-09-22 at 04:19 +0000, zhaolei@sourceware.org wrote:
> commit 707e14f086aa2b527f1e75aafdf6331356675e0a
> Author: Zhaolei <zhaolei@cn.fujitsu.com>
> Date:   Mon Sep 22 12:15:11 2008 +0800
> 
>     socket.stp (socket.aio_read/write): Fix the semantic error caused by the difference of kernel versions.

You are using the following logic to detect the right kernel version:

> + *
> + * 2.6.9~2.6.15:
> + *     static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf, size
> + * 2.6.16~2.6.18:
> + *     static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf, size
> + * 2.6.19~2.6.26:
> + *     static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov
>   */
>  probe socket.aio_read = kernel.function ("sock_aio_read")
>  {
>         name = "socket.aio_read"
>         _sock = _get_sock_addr ($iocb->ki_filp)
> -%( kernel_v < "2.6.19" %?
> -        size = $count
> +%( kernel_v <= "2.6.15" %?
> +       size = $size
>  %:
> -        size = _get_sock_size ($iov, $nr_segs)
> +       %( kernel_v <= "2.6.18" %?
> +               size = $count
> +       %:
> +               size = _get_sock_size ($iov, $nr_segs)
> +       %)
>  %)

Are you sure you want checks like kernel_v <= "2.6.18"? That would
include 2.6.18, but not 2.6.18.1. It seems better (but I don't have the
kernels here to test) to use kernel_v < "2.6.19" there and in other
places.

Cheers,

Mark

Re: [SCM] systemtap: system-wide probe/trace tool branch, master,updated. release-0.7-276-g707e14f

[Zhaolei]

Hello, Mark

> Are you sure you want checks like kernel_v <= "2.6.18"? That would
> include 2.6.18, but not 2.6.18.1. It seems better (but I don't have the
> kernels here to test) to use kernel_v < "2.6.19" there and in other
> places.
Thanks for your advice.
I checked sock_aio_read in 2.6.18.8, it is same as 2.6.18.
You are right, I will fix it soon, thank you!

B.R.
Zhaolei