Date:	Wed, 24 Sep 2008 12:13:47 -0400
From:	Mathieu Desnoyers <compudj@...stal.dyndns.org>
To:	Martin Bligh <mbligh@...gle.com>
Cc:	Peter Zijlstra <peterz@...radead.org>,
	Steven Rostedt <rostedt@...dmis.org>,
	linux-kernel@...r.kernel.org, Ingo Molnar <mingo@...e.hu>,
	Thomas Gleixner <tglx@...utronix.de>,
	Andrew Morton <akpm@...ux-foundation.org>,
	prasad@...ux.vnet.ibm.com,
	Linus Torvalds <torvalds@...ux-foundation.org>,
	"Frank Ch. Eigler" <fche@...hat.com>,
	David Wilder <dwilder@...ibm.com>, hch@....de,
	Tom Zanussi <zanussi@...cast.net>,
	Steven Rostedt <srostedt@...hat.com>
Subject: Re: [RFC PATCH 1/3] Unified trace buffer

* Martin Bligh (mbligh@...gle.com) wrote:
> Thanks for creating this so quickly ;-)
> 
> >> We can record either the fast way of reserving a part of the buffer:
> >>
> >> event = ring_buffer_lock_reserve(buffer, event_id, length, &flags);
> >> event->data = record_this_data;
> >> ring_buffer_unlock_commit(buffer, event, flags);
> >
> > This can, in generic, not work. Due to the simple fact that we might
> > straddle a page boundary. Therefore I think its best to limit our self
> > to the write interface below, so that it can handle that.
> 
> I'm not sure why this is any harder to deal with in write, than it is
> in reserve? We should be able to make reserve handle this just
> as well?
> 
> If you use write rather than reserve, you have to copy all the data
> twice for every event.
> 

I think we all agree that a supplementary copy is no wanted, but I think
this question is orthogonal to having a write wrapper. The way we can do
both is by using reserve/commit to deal with space reservation, and a
write() to perform the actual data write into the buffers once space has
been reserved.

Reserve/commit would allocate a variable-sized "slot" into the buffer.
We reserve X amount of bytes, and it returns the offset from the
buffer start where the allocated slot is. This reserve/commit mechanism
deals with synchronization (cli/spinlock or cmpxchg_local scheme...).

We can then use this offset to see in which page(s) we have to write.
This offset + len can in fact cross multiple page boundaries.

Doing this elegantly could involve a page array that would represent the
buffer data :

struct page **buffer;

And be given as parameter to the read() and write() methods, which would
deal with page-crossing.

e.g.

/*
 * Perform an aligned write of the input data into the buffer.
 *
 * buffer : page pointer array
 * woffset : offset in the page pointer array where write starts from
 * data : input data
 * len : length of data to copy
 *
 * Note : if a NULL buffer is passed, no copy is performed, but the
 * alignment and offset calculation is done. Useful to calculate the
 * size to reserve.
 *
 * return : length written
 */
size_t write(struct page **buffer, size_t woffset, void *data, size_t len);

Therefore, we could have code which writes in the buffers, without extra
copy, and without using vmap, in multiple writes for a single event,
which would deal with data alignment, e.g. :

size_t woffset, evsize = 0;

evsize += write(NULL, evsize, &var1, sizeof(var1));
evsize += write(NULL, evsize, &var2, sizeof(var2));
evsize += write(NULL, evsize, &var3, sizeof(var3));

woffset = reserve(..., evsize);

woffset += write(buffer, woffset, &var1, sizeof(var1));
woffset += write(buffer, woffset, &var2, sizeof(var2));
woffset += write(buffer, woffset, &var3, sizeof(var3));

commit(..., evsize);

Does that make sense ?

Mathieu

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists