lkml.org 
[lkml]   [2008]   [Sep]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v6] Unified trace buffer

On Fri, 26 Sep 2008, Martin Bligh wrote:

> On Fri, Sep 26, 2008 at 11:59 AM, Peter Zijlstra <peterz@infradead.org> wrote:
> > On Fri, 2008-09-26 at 14:05 -0400, Steven Rostedt wrote:
> >
> >> +struct buffer_page {
> >> + u64 time_stamp;
> >> + unsigned char body[];
> >> +};
> >> +
> >> +#define BUF_PAGE_SIZE (PAGE_SIZE - sizeof(u64))
> >
> > Since you're already using the page frame, you can stick this per page
> > timestamp in there as well, and get the full page for data.
> >
> > You can either use a struct page overlay like slob does, or add a u64 in
> > the union that contains struct {private, mapping}.
>
> What did you guys think of Mathieu's idea of sticking the buffer length
> in the header here, rather than using padding events? Seemed cleaner
> to me.

Actually I like the padding. This way when I move the event pointer
forward, I only need to compare it to a constant (PAGE_SIZE), or test to
see if the event is padding. Placing this into the buffer page, I will
have to always compare it to that pointer.

But I guess I could change it to that if needed. That doesn't affect the
API, as it is only internal.

I'm almost done with v7, perhaps I might try that with v8 to see if I like
it better.

-- Steve



\
 
 \ /
  Last update: 2008-09-26 21:55    [from the cache]
©2003-2014 Jasper Spaans. hosted at Digital Ocean